---
source_url: https://livrepository.liverpool.ac.uk/3194252/1/Output1_an-educators-guide-to-multimodal-learning-and-generative-ai.pdf
ingested: 2026-05-07
sha256: 43acf5a2885ed5d2dfce02556752648b5e7816972a4370ffed31d73f883c8a5a
---

Educators’ guide
to multimodal
learning and
Generative AI

The Original
Redbrick


An Educators’ Guide to Multimodal
Learning and Generative AI
Authors
Principal Investigators:
Tünde Varga-Atkins, University of Liverpool
Samuel Saunders, University of Liverpool

Co-investigators in alphabetical order:
Sue Beckingham, Sheffield Hallam University
Peter Hartley, Edge Hill University
Nayiri Keshishi, University of Surrey
Nataša Lacković, Lancaster University
Na Li, Xi’an Jiatong Liverpool University
Rob Lindsay, University of Liverpool
Run Wen, Xi’an Jiatong Liverpool University
Isabelle Winder, Bangor University

With contributions from our student partners:
Hadja Ba Ndiaye, University of Liverpool
Ayaat Jaway, University of Liverpool
Siyi Li, Xi’an Jiaotong Liverpool University
Yuan Yuan, Xi’an Jiaotong Liverpool University


Acknowledgements
We would like to offer our thanks to the following partners, who were instrumental in
helping the project team put this Guide together. First, to the Staff and Educational
Development Association (SEDA), both for funding to get the project off the ground, and
for the development of connections and for opportunities to publish interim updates on
the project as it progressed. Relatedly, we would like to offer SEDA Mentor, Professor Jennie
Winter, for offering her continuing support and advice on the project, and for organising
introductions with potential collaborators and for spotting dissemination opportunities.
We would also like to thank all our student research assistants, from both our partner
institution, Xi’an Jiaotong Liverpool University (XJTLU) and the University of Liverpool, who
were all pivotal to the creation of this guide. Our students were key in helping with the
initial scoping literature review (and its associated write-up), as well as organising and
running focus groups and generally providing excellent student perspectives at our
project team meetings. We would also like to thank Tianyu Zhang of XJTLU for their help
with supporting our student research assistants.
Thanks also to those who authored and submitted their Case Studies for incorporation
into this guide as examples of GenAI-enabled multimodal practice. These have helped
develop both the advice and ideas we are presenting in this guide, and provide useful
examples of it in action for readers to contextualise.
We would like to thank everyone who participated in one of our focus groups and survey
who provided us with a rich qualitative dataset on which to base our recommendations,
and those who participated in our Beyond Text: Generative AI in Multimodal Learning,
Teaching and Assessment Symposium in January 2025.

2


1. Introduction
This guide presents strategies and techniques for the incorporation of Generative AI
(GenAI) in multimodal forms of teaching, learning, and assessment. It focuses on the
intersection of multimodal learning and Generative AI (GenAI) and is written for both
educators and educational developers, who play a pivotal role in modelling pedagogical
innovation and guiding colleagues through evolving landscapes of teaching and learning.
The guide is a product of a 2024/25 SEDA Small Grants project, conducted between
academics and researchers spread across multiple higher education institutions across
Britain and beyond. The project drew on our own experiences with Generative AI in
multimodal contexts, as well as data from an extensive literature review, a case-study
collection exercise, a survey, and a suite of focus groups with educational developers,
educators, and students conducted during a collaborative research project in 2024–25.
At this initial stage, it is important to outline our position on the presence and use of GenAI
in higher education. Whilst ‘Artificial Intelligence’ has existed in different forms for decades,
the appearance and accessibility of GenAI since the release of the first iteration of
OpenAI’s ChatGPT in 2022 have enhanced its usability in learning and teaching contexts,
and has solidified GenAI as an emerging frontier in the application of AI in education
(Feng et al., 2025).
However, in this Guide, we opt for a pragmatic approach to the technology. We are not
uncritical ‘techno-fixers’ or over-enthusiastic adopters of GenAI, and nor do we view
Generative AI technology as the ultimate destroyer of higher education. Rather, and as
Vallor (2024) eloquently puts it, we contend that AI is a ‘mirror machine’ that reflects
ourselves back at us, and, just like Vallor, we wish to provide educators, educational
developers and students with an approach that gives them:
“a middle way between passive resignation to AI technology as a replacement for
human agency and rejecting AI as an existential threat that must be opposed and
defeated” (Vallor, 2024).
Part of the challenge of examining GenAI in educational contexts is the tension between
its speed of development and evolution, and its already-entrenched position in
pedagogy, as well as its endorsement by national and educational institutions. On the one
hand, even as we write this guide, GenAI is advancing rapidly – indeed, in the final stages
of editing, OpenAI released GPT-5, the very latest model embedded in the company’s
flagship ChatGPT platform, and which supposedly represents a significant step forward.
On the other hand, Generative AI is already ‘here to stay’; students, staff, professionals,
lay-users, and the public all use Generative AI in their daily lives – sometimes without
realising it. Simply ignoring the reality of GenAI – or, even worse, actively rejecting that
reality – and its potential impact on universities and higher education, will not help
3


students understand the machines better, nor will it help them to make better choices
about what technologies to use, and when.
However, we should state that we do not believe that talking about Generative AI is the
same as uncritically endorsing the use of Generative AI. Again, we do not necessarily
endorse Generative AI in educational contexts, and we do not argue that it should be used
at every opportunity. Rather, we contend that wilfully ignoring it and/or actively rejecting it
does students a disservice, and will ultimately make them more likely to misunderstand
the technology, either accidentally or deliberately, in future contexts. As a result, educators
can, and should, support students in critical about their GenAI use.
Digital technologies have both intensified and shone a light on the inherently multimodal
nature of modern society — where learning and communication extend beyond text and
speech to include the visual, aural, or tactile (Kress, 2009; Cope & Kalantzis, 2024). As
Generative AI technology becomes increasingly embedded in everyday life, it is vital for
higher education to design multimodal learning, teaching, and assessment approaches
that foster students’ ability to think both critically and effectively in a digitally mediated
context (Lim, 2020; Lacković, 2020; Li et al., 2024; Zawacki-Richter et al., 2019). Hitherto, the
discourse on GenAI has focused on concerns around maintaining academic integrity,
critiquing AI, or unquestioningly celebrating or actively promoting its integration in
education spaces. Indeed, there have been calls to fundamentally and entirely re-align
our assessment practices for the era of GenAI (Furze, 2024). In this guide, we turn our
attention to understanding GenAI’s multimodal pedagogic potential, whilst remaining
critical and cautious in our appraisal and suggestions for GenAI-enabled multimodal
learning designs.
In a multimodal learning environment, educators aim to move beyond text-dominant
formats, and instead incorporate and combine a variety of communication forms to
better support student learning. Digital multimodal learning involves designing learning
activities that draw on multimodal artefacts and utilise a combination of ‘semiotic
technologies’ (see Lim & Tan-Chia, 2023, Lacković, 2020, and Lacković & Olteanu, 2023).
These technologies help to combine image, sound, touch and other modes into
multimodal artefacts that are better than single-mode artefacts for meaning-making
(Kress, 2009; van Leeuwen, 2017), commonly a combination of verbal and some other
mode of expression.
Examples of digital multimodal artefacts might include infographics, digital posters,
videos, presentations, virtual simulations or any visual media applied in an educational
dialogic context. A key aspect of multimodality is the way that different modes are
assembled (e.g. text and image in an infographic or in an inquiry graphic (Lacković,
2020)). This assembly (by machines, producers, students or educators) provides extra

4


meaning beyond the sum of its parts, and informs interactions and discussions in the
higher education context.

1.1 What do we mean by Generative AI, and how it is developing?
Before proceeding, it is perhaps useful to explain what we mean by Generative AI. In fact,
this is a useful question to discuss with colleagues and students before using it any way,
as there are many common misconceptions. For example, there is a common belief that
GenAI technology can function efficiently as a ‘search engine’, while, in fact, it is
notoriously unreliable as a means of searching for information, given that it prioritises
relevance based on the prompt over accuracy or impartiality of information retrieved
(Hemsworth et al., 2024).
GenAI employs deep machine learning techniques to process information contained
within huge datasets to generate outputs based on human prompts inputted by the user.
Most GenAI we refer to are Large Language Models (LLMs) trained on vast amounts of text
data to decode, generate, and manipulate human language. However, GenAI is
increasingly capable of producing multimodal content, including (but not limited to) text,
speech, audio, image, video and even three-dimensional models (Fui-Hoon Nah et al.,
2023). GenAI technology enables users to create, manipulate, and adapt content and
integrate different semiotic forms to produce multimodal artefacts, and thus can be
embedded into pedagogical practices that already emphasise diverse modes of
engagement.
GenAI’s rapid development has been accompanied by suggestions on how to define and
use this technology. For example, Mollick suggests we should consider it to be a ‘cointelligence’ (Mollick, 2024), while Cope & Kalantzis (2024) suggest that we should
understand it as a unique ‘co-creator’ that works alongside users in an assistive but
unique role. They call this cyber-social learning – a collaborative partnership between
human and machine intelligences, each with distinct, but complementary, strengths for
completing an activity (2024; see also Galla et al., 2025). They suggest that this
collaboration ‘enables new processes for knowledge creation’, where educators and
students learn by evaluating, refining and re-imagining AI outputs, assembling them into
multimodal artefacts. In other cyber-social contexts (Galla et al., 2025), we have also
recently seen several LLMs claiming the ability to engage in ‘Socratic dialogue’ with users
(Leoste et al., 2025).
However, while human and artificial intelligences can work together and have a unique
role to play within a cyber-social partnership, the term ‘intelligence’, taken as part of
‘Generative AI’ can, and perhaps should, be challenged in favour of more specific
computer-science-driven terminologies, such as LLMs (large language models).
‘Intelligence’ implies ‘consciousness’ that AI simply does not have, despite its (and their

5


parent companies’) attempts to lead users into believing it does. Debating whether GenAI
can legitimately be described as ‘intelligent’ is beyond the scope of this guide, but it is an
important consideration when thinking about LLMs and their position in the learning
process – particularly in cyber-social contexts. Indeed, it is GenAI’s very lack of
‘intelligence’, either emotional or intellectual, that highlights how it operates in the cybersocial relationship with a human, whereby both parties occupy unique but symbiotic
positions and consequently complement each other. While GenAI can generate and
transform multimodal content at scale and with (sometimes extreme) efficiency, it lacks
understanding, spontaneity, interpretation, emotional nuance, or critical and ethical
judgement – the very qualities that characterise human intelligence. In other words, GenAI
can offer some raw material – drafts, artefacts and prototypes - but educators and
learners are needed to bring vision, purpose, nuanced critique and meaning-making.
The guide’s aim is to encapsulate strategies for the effective incorporation of GenAI in
multimodal teaching, learning, and assessment and to position Generative AI most
effectively within the cyber-social relationship with human users to maximise the
potential of the technology to produce, interpret and engage with multimodal artefacts
and thereby improve meaning-making. We therefore interpret ‘multimodal GenAI’ in all
these different ways:
•

Multimodal GenAI can refer to platform capabilities that utilise modalities beyond
text-to-text (e.g. text-to-image, text-to-speech, text-to-video, speech-to-text
etc.), or;

•

We can refer to a multimodal learning or teaching activity itself (e.g. a lecture or a
virtual simulation) that utilises GenAI within its process (whether GenAI itself is textto-text or multimodal), or;

•

Using GenAI to convert one artefact/modality (e.g. slides or images) into another
modality (e.g. text or sound).

6


Figure 1. The multimodality continuum of learning modalities and experiences supported by GenAI
(from text, visual text, audio/video, interactive audio/video, avatars and immersive simulations).

This guide offers broad principles and approaches rather than platform-specific
suggestions to ensure relevance across disciplines and learning contexts. This was largely
necessary; even during the lifespan of this eighteen-month project (2024/25), GenAI’s
multimodal capabilities have evolved so rapidly that listing very specific concrete
examples, using specific apps or platforms, risks the information becoming quickly
outdated. Indeed, Figure 1 illustrates GenAI capabilities’ development in terms of
educators’ uses of multimodal resources from text to immersive simulation. We might
contend that GenAI currently offers interactive content to learners and educators, via
real-time interactivity with avatars or personas. However, just a year ago, this would
perhaps have been closer to static textual, or perhaps audio/visual, content.

1.2 What do we mean by multimodality, key terms and links to
GenAI?
As this guide is about multimodal learning, it is worth briefly outlining what we mean by
‘multimodality’ in the context of this guide and, particularly, in relation to its connection
7


with Generative AI (GenAI) technology. We also offer a few key terms, building on the
MODE project (see below), that one might expect to see when reading about multimodal
pedagogies and some useful definitions for them.
Gunther Kress’s seminal 2009 book “Multimodality: A Social Semiotic Approach to
Contemporary Communication” provides the basis for this guide’s understanding of
multimodality. Kress contends that forms of communication are, inherently social, driven
by the context in which they appear to help observers to make meaning, and that
artefacts that combine multiple semiotic forms – text, image, sound, touch, smell, etc. –
create greater degrees of expression and meaning in both communicator and
communicatee than single-mode artefacts (2009).
This is relevant in the context of the appearance and proliferation of GenAI, given the
different forms of information that it can create. GenAI has the potential to fundamentally
transform the way we both create content such as text, image, video and sound, but also
the way we combine these forms together into new artefacts that work to develop a
greater understanding of a given topic, or deeper meaning concerning it, than either
single-mode artefacts or entirely human-created ones. This is not without its dangers, as
the content created can serve all sorts of purposes, even identify theft, when someone’s
facial features are used to create fake videos or images. GenAI might be able to create
and even combine semiotic forms, but the decisions about what to combine, where to
place it, size, shape, volume, etc., are all human choices, which helps to exemplify the
symbiotic relationship between the two intelligences – human and artificial, or coined as
‘cyber-social’ by Cope and Kalantzis (2024).
To define multimodality further, there are specific terms associated with this theory worth
outlining here, particularly in terms of their relationship to GenAI. The following
explanations either use or adapt the exact wording of the ‘Glossary of Multimodal Terms’.
You can use this resource for further information on multimodality terminology,
information and authors’ details, but here are some of the key terms we make use of
throughout this guide:

8


Mode (multimodality) of GenAI: Mode classifies a ‘channel’ of representation or
communication for which previously no overarching name had been proposed (Kress &
van Leeuwen, 2020). Examples of modes include writing and image on the page,
extending to moving image and sound on the screen, speech, gesture, gaze and posture
in embodied interaction. GenAI inputs or outputs can be multimodal as they consist of
more than one mode, but it can be said that all communication/expression is multimodal.
Multimodality can also be conceived in terms of the difference of modes between inputs
and outputs (e.g. text-to-image, image-to-text).

Multimodal Affordance of GenAI: Adapted from Gibson’s (2014) theory of affordances by
Kress (2009), the term ‘modal affordance’ has particular currency in multimodality. It
refers to the potentialities and constraints of different modes – what it is possible to
express and represent or communicate easily with the resources of a mode, and what is
less straightforward or even impossible – and this is subject to constant change in
environments or interactions. With regards to GenAI and learning, it refers to the various
ways in which meaning is made with GenAI to form part of teaching, learning and
assessment. It invites educators and learners to consider how GenAI makes meaning from
its database, training and through various prompts. It also means that we need to
consider what potential and challenges there are for reflection and critical thinking when
learners engage with GenAI and its outputs. GenAI affordance is neither something that
GenAI contains, nor does it refer to perceiving it (perception) – it is somewhere inbetween as an evolving potential in between what the technology offers and whether and
how this is perceived. It is about how GenAI can function in a learning context and what
activities it can be part of, depending on users and the platform. In a basic example, we
might argue that text-to-image GenAI platforms (e.g. DALL-e) have an ‘affordance’ to
create images from the inserted prompt words. This affordance can be used differently in
learning activities and for learning.

Multimodal discourse of GenAI: All multimodal texts, artefacts and communicative events
are always discursively shaped; all modes, in different ways, offer means for the
expression of discourses. For example, GenAI representations have been found to be
biased in different ways, as aligned with the broader discourse that is circulating on the
internet as well as human data and training input. What this means is that learners need
to understand that the data used to train GenAI carries a specific discourse that may
reinforce certain hierarchies and power dynamics in societies. This is why it is important to
discuss GenAI production with students critically.

9


1.3 Structure and how to use this guide
As we have said, we fully recognise the importance of responsible use of GenAI. We are
aware that frivolous, uncritical, unquestioned and ignorant use of the technology is, at
best, unlikely to be of much use to the user in producing the response they want, and at
worst both irresponsible and in some cases dangerous. Consequently, we initially provide
a section on the responsible and ethical use of GenAI in multimodal educational contexts
(Section 2), before then summarising some perspectives on developing GenAI literacy in
both students and staff in a multimodal educational context (Section 3).
Once we have established these important contextual perspectives, we then move to offer
practical and applicable learning designs for integrating GenAI tools for multimodal
learning. The guide is organised into three strands: Teaching (Section 4), i.e. how GenAI
can be used to represent subject knowledge multimodally and create activities/resources
for students; Learning (Section 5) i.e. how students could encounter, explore, evaluate,
critique and express ideas and artefacts via multimodal GenAI to improve their knowledge
and skillsets; and Assessment and Feedback (Section 6), i.e. how students or educators
can use GenAI to create or critique/reflect on multimodal artefacts in formative or
summative assessment capacities, or use the technology to create or engage with
feedback in multimodal ways, or else self-reflect using conversational Generative AI.
Each of the three sections contains sub-sections on what it is, why do it and how the idea
or practice might work in-situ, as well as further ideas and selected cases studies on how
educators might put some of these ideas into practice for themselves. We gathered these
case studies from across the sector as a data-gathering exercise to determine what kinds
of practices were already occurring, and we have included all the case studies in full as a
separate Appendix (Appendix A) at the end of this guide. Section 7 includes some advice
and considerations for choosing GenAI platforms when designing multimodal GenAI
activities. Finally, Section 8 presents a unifying model for designing multimodal learning
with GenAI.

10


2. Responsible use of AI in a multimodal context
Despite the way GenAI is often marketed as efficient, trustworthy, innovative and
invaluable to boosting users’ productivity, integrating GenAI into multimodal learning
brings substantial pedagogical, social and ethical challenges. The responsible use of
multimodal GenAI in higher education requires a careful balance of ethical, pedagogical,
legal and practical factors. Concerns include AI’s accuracy, reflection of entrenched social
bias(es), and the potential for over-reliance, which may weaken student voice or diminish
their capacity for critical thinking. Additionally, some educators and students feel
uncertain about how to use GenAI tools effectively, ethically, or correctly within the context
of (sometimes vague or ambiguous) institutional policy. Ensuring AI-enhanced teaching
is informed, equitable, and values-driven requires deliberate strategies and a strong
foundation in AI literacy (Saunders et al., 2024).
These issues can potentially be addressed by employing curriculum design(s) that
foreground the development of AI literacy and criticality. Another general strategy is to
introduce the complexities of the ‘costs’ of AI and GenAI in both environmental and social
contexts – issues which are often ignored or examined only superficially in publications
from GenAI ‘enthusiasts’, and which are often concealed by companies that produce
Generative AI software. A useful starting point for educators seeking to take this approach
is Beckingham and Hartley’s (2025a) article that suggests four areas to consider when
looking at the cost of using GenAI: cost to the individual, cost to the environment, cost to
knowledge and cost to future jobs. In each of these areas, we can consider both
quantitative and qualitative factors; for example, individual costs include the quantitative
costs of subscribing to different apps, alongside more qualitative factors such as the
impact of Generative AI on users’ mental health.

Figure 2. Cost of Generative AI to the individual, environment, knowledge and future jobs

Discussion of each, or a combination of, these areas can generate action plans for both
staff and students to develop their understanding of using GenAI responsibly, as the
following examples illustrate.
11


2.1 Costs to the Individual/Individual Responsibility
As ‘accountability’ is an important theme here, we can begin by promoting clarity on what
is acceptable use and transparent authorship. Assessment design, for example, should
make explicit when GenAI can and cannot be used. Providing a pedagogical purpose for
using GenAI needs to be scaffolded. Students need to learn how it can be used to
enhance learning. Tools should be selected with purpose based on the specific learning
outcomes they enhance. It is important to recognise that every GenAI will be biased in its
outputs and is therefore likely to amplify stereotypes, misrepresent information or be
inaccurate – and it is rare, if not unprecedented, for a Generative AI platform to recognise
that it has done so, let alone inform the user. Given that the platform has no sense of
social justice or any kind of consciousness, it cannot know when it has reinforced an
existing prejudice or biased perspective. Indeed, it cannot ‘know’ anything. Consequently,
we need to encourage students to engage critically with outputs from GenAI rather than
passively consume them and assume they are simply correct and free from prejudices.
From a legal perspective, privacy, data protection and consent should also be considered
with respect to what is uploaded to and how this data may be used by GenAI platforms,
e.g. for further training.
A further consideration is equity of access. There are an enormous number of GenAI tools
now available, with different specific purposes. There is also a growing trend for GenAI to
be embedded into existing technologies or applications, opening new ways of working
with existing tech. But many GenAI apps carry a financial cost, while embedded GenAI is
often placed behind a ‘premium tier’ paywall, disallowing access to either the app, or to
certain features of it, to users unless they are willing to pay extra for it. Consequently,
educators need to work to ensure all students have an equitable experience; whilst free
access to some tools is available, this is constantly changing. Having an approved tools
list of recommended GenAI platforms can be helpful in this regard, but this does need to
be regularly reviewed.

2.2 Costs to the environment
There is mounting evidence that using GenAI – especially generating images, video and
audio (or, particularly, artefacts that combine these) – requires significantly more
computational resources and energy compared to pure text-based generation. A fruitful
study by Luccioni et al. (2024) highlights this, and she is developing an AI energy score
model that will guide individuals in choosing models (and ways to run them) for different
tasks based on their energy efficiency. Unfortunately, and as Luccioni et al. note, precise
information on the exact energy costs of different GenAI platforms is very difficult to
extract from the producers and suppliers (2024). This makes it very difficult to devise clear
guidelines on responsible use based on the amount of GenAI use, although it does provide
12


a solid basis to confirm that their energy use is generally substantial and should, at least,
be considered by all users before choosing to engage with a GenAI platform.

2.3 Costs to knowledge
GenAI has the potential to help students to acquire knowledge and develop their critical
capacity, although this tends to be short-term. That said, there are also a significant
number of studies which demonstrate that over-reliance on GenAI can potentially have a
negative or damaging impact on the ways students acquire knowledge as it removes the
process of sourcing that knowledge that is often integral to retention, as well as retention
itself (Yan et al., 2025). Indeed, a recent study from MIT demonstrated that students who
start by clarifying their own understanding before they consult GenAI can experience
major benefits (Naughton, 2025).

2.4 Costs to future jobs
In 2025, some employers specifically attributed decisions on reducing staff numbers to
the growth of Generative AI. For example, the CEO of Anthropic warned that ‘AI could
eliminate half of all entry-level white-collar jobs’ (Morris, 2025). In addition, several
research reports also appeared, suggesting that particular job roles or tasks were under
threat from GenAI. To help make sense of this, Tomlinson et al. (2025) devised an ‘AI
applicability score’ for specific occupational roles. The higher the score, the more
vulnerability to GenAI, and they ultimately found that:
“the highest AI applicability score for knowledge work occupation groups such as
computer and mathematical, and office and administrative support, as well as
occupations such as sales whose work activities involve providing and
communicating information” (Tomlinson et al., 2025).

2.5 Implications
The main implication for higher education in relation to any use of Generative AI is to
encourage users to engage in critical reflection in both the context of starting the task and
the context of completing it – users should consider asking themselves: is the use of GenAI
necessary for the task? Would other ‘less-damaging’ alternatives suffice? If I ‘out-source’
this task to GenAI, does this help me develop my own capacities and abilities?
Reducing the cost of using GenAI can also be approached through small but manageable
microtasks by students, educators and institutions. These can help to build awareness,
reduce unnecessary usage, and promote more energy-efficient practices. Consider the
following suggestions (Welsh & Milne, 2025):

13


•

Use GenAI mindfully – ask ‘do I really need GenAI for this task?’
o

Consider using the technology for idea generation or creative reflection. For
example, if you need an image, could royalty-free sites like Pixabay or
Pexels that offer free-from-copyright images serve your purpose better?

•

Limit iterations
o

Work to provide GenAI platforms with clear and detailed prompts in the first
instance to get the result you want/need more quickly, instead of excessive
‘trial and error’ multiple attempts that use up more energy.

•

Digital decluttering
o

Delete unwanted generated images and text. Download and store only the
final versions.

•

Run GenAI demonstrations as a group activity rather than each student generating
outputs individually.

14


3. Developing AI literacy in a multimodal context
In the previous section, we emphasised the importance of the critical and responsible use
of GenAI in education. This tied to developing AI literacy, enabling educators and learners
to /question) GenAI in a multimodal learning context. Indeed, developing AI literac(ies) is
of critical importance in the context of either engaging or disengaging from the
technology, because if one does not understand the platforms, their functions, contexts,
impacts and perceptions, one cannot justifiably use it, or justifiably reject it.
Below are some strategies to consider in developing your learners’ GenAI literacy, whether
they are staff or students. It should be noted that these are not some ultimate definitions
of AI literacy in multimodal learning, as the terms ‘literacy’ and ‘literacies’ have been
interpreted in myriad ways. Rather, these are some basic approaches that may help
educational developers and educators begin fostering GenAI literacy in their staff and
student audiences.

3.1 Different levels of AI Literacy in multimodal learning
There are several ‘levels’ of GenAI literacy, which can be characterised by particular
activities or abilities in users. However, we should say that this is not an exhaustive or
comprehensive list, but rather a basic benchmark for identifying differing levels of literacy
with the technology that may be useful as a starting point:
•

Basic literacy: Awareness of multimodal GenAI platforms, their capabilities, and
appropriate uses in educational context (e.g., creating prompts, generating visual
outputs).

•

Intermediate literacy: Ability to co-create multimodal content, critically evaluate
multimodal AI outputs, and scaffold uses (e.g., transforming lecture notes into
visuals or podcasts).

•

Advanced literacy: Designing activities or assessments that incorporate
multimodal uses of GenAI, fostering critical analysis and engagement, and leading
ethical and philosophical discussions on AI implications in academia and wider
contexts.

3.2 Key teaching competencies for GenAI in multimodal learning
Similarly, there are some key teaching competencies and activities that can perhaps be
used in a classroom setting that can help develop AI literacy in-situ:
•

Scaffolded prompting: Advising students how to craft and iterate on prompts to
refine multimodal GenAI outputs (e.g., generating diagrams, podcast scripts, or
video summaries) – if possible, with help from dedicated technology experts.

15


•

Evaluation frameworks: Encouraging students to critique GenAI content for
accuracy, bias, and coherence. For instance, having students examine and
annotate AI-generated essays or visual outputs.

•

Ethical protocols: Establishing clear boundaries on acceptable use (e.g., ‘do not
use AI to write reflections’; ‘do use it for brainstorming visuals’) to ensure academic
integrity and data protection.

3.3 Strategies for developing AI Literacy across different levels
3.3.1 Individual Level
•

Students: Participate in workshops on GenAI use in creative multimodal tasks,
practice prompt crafting, and document AI usage in reflective assignments.

•

Educators: Engage in professional learning, experiment with low-stakes uses of
GenAI, and reflect on ethical practices.

•

Educational Developers: Curate examples, develop guidelines, and coach staff on
GenAI-enabled pedagogies.

3.3.2 Module Level
•

Embed GenAI literacy into learning outcomes (e.g., ‘critically evaluate AI-generated
design concepts’).

•

Offer optional multimodal tasks that include GenAI use with clear rubrics and
support.

•

Include creative and reflective components where students analyse and critique
their own, or others’ GenAI use.

3.3.3 Programme Level
•

Develop cross-module policies and examples on GenAI use.

•

Promote consistency and transparency in how AI-integrated tasks are introduced
and assessed via workshops, cross-staff activities and discussion tasks.

•

Align GenAI practices with graduate attributes such as criticality, creativity, and
digital fluency.

•

Consider embedding AI literacies on academic skills and/or literacies modules
shared across programmes within departments or schools.

3.3.4 Institutional Level
•

Provide clear policies on GenAI use in learning and teaching, with checklists for
permissible uses.

•

Offer vetted tools and ensure data privacy protocols are enforced.

16


•

Create professional development pathways, support communities of practice (e.g.,
AI innovators groups), and showcase successful implementations to build
confidence and capability.

•

Offer platforms for debate and critique on the role and uses of GenAI in education.

17


4. Multimodal Generative AI in Teaching
Generative AI (GenAI) tools significantly broaden the possibilities for multimodal teaching
by enabling educators to design, adapt, and deliver content across a range of media
(Bond et al., 2024; Yun et al., 2025). These technologies support not only the creation of
multimodal artefacts such as images, audio, video, and interactive simulations, but also
the transposition of content from one format to another. For instance, GenAI might be
used to convert lecture notes into podcasts, diagrams into explanatory narratives, or
seminar discussions into visual summaries (Prinsloo et al., 2024). This ability to work fluidly
across modes introduces new opportunities for educators to make complex or abstract
concepts more tangible, and to design learning experiences that better accommodate
the diversity of students’ learning preferences and needs (Fui-Hoon Nah et al., 2023).
In a multimodal teaching-learning environment, GenAI allows educators to move
decisively beyond text-dominant approaches. Text-to-image and text-to-video tools can
generate diagrams, animations, and explanatory visuals in minutes; audio synthesis can
create narrated explainers or podcasts to accompany slides; and large language models
can rapidly produce draft quiz questions, scenario descriptions, or alternative
explanations pitched at different levels of complexity (Koppam et al., 2024). By embedding
these outputs within teaching activities, educators can enrich their teaching repertoire
while saving time on labour-intensive production tasks (Alasadi & Baiz, 2024).
The role of GenAI in this context is best understood as amplification rather than
substitution; it does not replace the pedagogical expertise or creativity of educators, but
instead augments their capacity to respond flexibly to students’ needs (Walter, 2024; Hall,
2024). Educators can focus on higher-order aspects of teaching such as sequencing,
framing, and critical discussion (Tay et al., 2025), while delegating lower-level production
tasks to AI (Kim et al., 2025). This shift enables more responsive and iterative teaching
design, where materials can be refined in real time to reflect emerging learning needs or
student feedback (Li et al., 2024).
At the same time, embedding GenAI into multimodal teaching demands a critical and
ethical orientation. AI-generated outputs are shaped by the datasets on which they are
trained, meaning they may reproduce biases, inaccuracies, or stereotypes (Pinski &
Benlian, 2024). Educators therefore have a dual responsibility: to model the critical
interrogation of AI outputs within their own teaching practice, and to scaffold students’
ability to do the same. This is especially vital where AI is used to represent disciplinary
knowledge visually or narratively, as inaccuracies can be less immediately visible than in
text (Chen et al., 2023).
Considerations of equity and sustainability must also inform multimodal AI use. Access to
high-quality AI tools is uneven across institutions and student populations, with premium

18


features often gated behind paywalls. Image and video generation carry significant
computational and environmental costs, so purposeful adoption is essential, ensuring
that multimodal AI is used where it genuinely enhances learning rather than as a novelty
(Bolick & da Silva, 2024).
Integrating GenAI into multimodal teaching also creates opportunities to develop futurefacing skills. As workplaces increasingly demand fluency in interpreting, critiquing, and
collaborating with AI-generated content, students benefit from encountering these
practices within their studies (Hsiao & Zhang, 2023). Educators who intentionally model
transparent AI use, explaining prompts, evaluating outputs, and reflecting on limitations,
can nurture these literacies while simultaneously enriching disciplinary teaching (Ng et al.,
2024).
This section explores how educators can use Generative AI (GenAI) to support and
enhance multimodal teaching practices across diverse disciplines in higher education.
Multimodal teaching involves engaging students through multiple forms of media—such
as text, visuals, audio, video, and interactive elements—to better support different learning
styles, deepen understanding, and foster creative expression. GenAI offers a powerful set
of tools that can facilitate the design, delivery, and co-creation of such multimodal
content, even for educators or students with limited technical or design expertise. The
section provides concrete examples of how GenAI can be used to create visual materials,
assist in the delivery of engaging content, and support collaborative and creative learning
processes. It also highlights ways in which GenAI can be integrated into teaching activities
that explicitly promote AI literacy and critical thinking, as well as strategies for co-creating
content with students to foster participatory learning environments.
In addition to practical applications, the section critically examines both the opportunities
and limitations of using GenAI in education. It discusses the potential for GenAI to enhance
pedagogy, lower barriers to multimodal creation, and increase student engagement and
creativity, while also addressing risks such as over-reliance on AI, the loss of reflective
learning processes, and concerns around ethics, sustainability, and academic integrity. To
support thoughtful and responsible use, the section offers practical strategies and
pedagogical tips for integrating GenAI into multimodal teaching. These include
approaches for saving time, modelling AI use transparently, embedding ethical guidelines,
designing for human–AI collaboration, and treating AI literacy as a core graduate
capability. Finally, the section features case studies that illustrate how educators in
different subject areas are using GenAI in innovative and contextually relevant ways.
These examples provide inspiration and insight into how AI can be meaningfully
embedded in teaching practice to enhance learning, creativity, and digital fluency in a
rapidly evolving educational landscape.

19


4.1 What is it? GenAI in multimodal teaching
In this section we offer examples of educators using GenAI to create or adapt teaching
materials.

4.1.1 Using AI for creating visual content
Educators might use Generative AI to:
•

Generate images to help produce diagrams, icons, or images/visuals for different
purposes, such as academic posters or presentations - this is especially useful for
analysing such GenAI products critically to enhance students’ critical awareness.

•

Create multimodal elements or icons to be used within infographics - using textto-image tools – from financial concepts or data, demonstrating understanding
visually.

•

Create visual metaphors using GenAI-generated images to represent abstract
concepts in relation to teaching context, combining visuals with reflection.

•

Generate short educational videos or animated explainers from written prompts,
making complex ideas explained in different modalities that are more experiential
and sensory, such as through motion graphics and narration.

Sample GenAI Prompt:
I’m a university lecturer preparing an academic poster on [insert topic, e.g., ‘neural
network architectures’ or ‘comparative political systems’]. Please suggest a list of AIgenerated image ideas - such as conceptual diagrams, icons, or illustrative visuals - that
can help convey the idea of [insert idea or message to be depicted in poster] visually,
minimising the need for dense text. The visuals should be intellectually rigorous, clear, and
appropriate for a higher education audience. Feel free to include suggested labels or
annotations.
(Note: for the prompt to work well, it may need more context and specificity to minimise
the requirement for iterative refinement.)

20


Case study summary: Teaching – Improving engagement with eLearning by creating
multimodal video-explainers by Vivien Shaw, Northern College of Acupuncture –
Biomedicine
In response to student feedback on long, hard-to-navigate eLearning materials, a series
of short, multimodal ‘explainer’ videos were created to link digital content more clearly
with in-person learning outcomes. Using tools like Microsoft Word’s transcribe function
and ChatGPT, she summarised hour-long online learning sessions, mapped them to
relevant learning outcomes, and generated slide decks. These were then developed into
10-minute narrated videos, illustrated with selected slides to support clarity and revision.
This approach, grounded in Universal Design for Learning, significantly reduced student
anxiety and helped clarify expectations for assessment. While some students engaged
less with the full online content, the explainers were highly valued for their accessibility
and structure. The model is easily transferable, using free or low-cost tools and adaptable
prompts, offering a practical way to enhance digital learning design and student
confidence.
(Full case study is included in Appendix A.)

4.1.2 Integrating GenAI into teaching delivery
Educators might use Generative AI to:
•

Deploy an AI-powered Q&A chatbot trained on course materials to provide
students with interactive, conversational text-based support.

•

Use AI to generate multimedia components for lectures, such as visuals, analogies,
or quiz questions, making sessions more engaging and develop particular skills
and competences.

•

Experiment with AI to create multimodal activity briefs or learning materials,
adding diagrammatic or audio explanations to traditional text.

21


Sample GenAI Prompt:
I am designing a university-level task or activity brief for a course on [insert subject, for
example, ‘environmental policy analysis’]. I want to create a multimodal brief that
includes a clear written description of the activity, together with a supporting diagram
and a short audio explanation to enhance student understanding.
Please help me:
•

Draft a clear, academically appropriate description of the activity.

•

Suggest a concept diagram or flowchart I can generate with an AI image tool to
visually represent the activity or workflow.

•

Write a 1–2-minute script I can use to create an audio explanation using a text-tospeech tool.

Case study summary: Teaching – Perk Up Your Academic Journey: A Coffee Shop
Adventure by Dr Laura Sharp, Eric Davies, Mia Wilson & Ailsa Foley, University of Glasgow –
Academic Values, Originality, and Plagiarism
‘Choices at the Coffee Shop’ is a gamified, AI-supported learning tool that engages
students with academic integrity through an interactive ‘choose your own adventure’
experience. Embedded in the University of Glasgow’s asynchronous module Academic
Values, Originality, and Plagiarism, the game presents real-world scenarios to help
students understand the consequences of their academic decisions in a safe, engaging
space. Using AI tools for image generation, animation, and voiceovers, the project team
created diverse characters and immersive experiences. Feedback shows strong student
endorsement, with 96% reporting improved understanding of plagiarism. The design
process involved collaboration across academic and digital teams, iterative feedback,
and careful handling of AI’s limitations in representing diversity. This accessible, costeffective approach demonstrates how AI can enhance active learning, support diverse
learners, and humanise complex topics in online education
(Full case study is included in Appendix A.)

22


4.1.3 Using GenAI in collaborative and creative processes
Educators might use Generative AI to:
•

Auto-generate ideas, documents, or simple images for brainstorming and design
thinking within collaborative digital environments (such as a Miro board).

Sample GenAI Prompt:
I’m facilitating a collaborative session with university students/researchers on a Miro
board for a [design sprint / group research project / seminar workshop] focused on
[insert topic, e.g., ‘climate justice frameworks’ or ‘AI ethics in healthcare’].
Please help generate:
1.

A set of conceptually rich ideas or discussion prompts related to the topic.

2.

Academic-style sticky notes or text snippets we can use to structure the board (e.g.,
themes, questions, frameworks).

3.

Suggestions for simple AI-generated images, diagrams, or icons to visually represent
theoretical models, key concepts, or tensions.

The output should support higher-order thinking and collaborative knowledge
construction at a university level.

Case study summary: Teaching – Peer Conflict Resolution in Group Work: Using GenAI
to Represent Subject Knowledge Multimodally by Rob Lindsay, University of Liverpool –
Interactive Group Work
This interactive activity helps students navigate common interpersonal conflicts in
university group work through realistic, AI-generated scenarios. The resource uses avatar
videos, synthesised voices, and branching dialogue to immerse students in challenging
situations—such as miscommunication or unequal workload. Teachers can facilitate
student learning about groupwork that helps explore different responses, receive
immediate feedback, and reflect on their choices, supporting the development of
communication, collaboration, and cultural competence. Designed using tools like
ChatGPT, ElevenLabs, Adobe Express, and H5P, the multimodal design supports authentic,
transferable skill development while enhancing engagement in asynchronous settings.
Educators can customise the content by tailoring the scenarios to suit their subjectspecific contexts.
(Full case study is included in Appendix A.)

23


4.1.4 Activities explicitly teaching AI literacy and criticality
Educators might use Generative AI to:
•

Design tasks where students use various multimodal AI tools (image identifiers,
audio identifiers, text generators) on a topic and then critically evaluate the
accuracy and quality of the AI outputs. This helps them understand AI’s limitations
and biases and develops evaluative judgement.

•

Design (potentially dialogic) activities focused on experimenting with prompt
variations to understand how prompt input affects multimodal outputs and bias.

•

Engage students in documenting and critically reflecting on their use and/or
rejection of use of AI.

Sample GenAI Prompt:
I’m designing a university-level activity where students will explore a topic of their choice
(e.g., climate change, medical misinformation, historical events) using a range of GenAI
tools - such as image recognition tools, audio transcription or identification tools, and text
generators.
Please help me:
1.

Draft a task brief that instructs students to use at least two different GenAI tools to
gather or generate information on their topic.

2.

Include prompts that guide students to critically evaluate the accuracy, relevance,
and potential biases in the GenAI-generated outputs.

3.

Suggest reflection questions that help students consider the capabilities and
limitations of AI, and how this impacts trust, validity, and academic use.

The task should support the development of evaluative judgement, digital literacy, and
critical thinking at a higher education level.

4.1.5 Co-create multimodal output/delivery
Educators might use Generative AI to:
•

Generate short video snippets from lecture content for VLEs.

•

Summarise documents and convert them into different formats for accessibility
(e.g., podcast scripts).

24


Sample GenAI Prompt:
I’m working with a university-level text/document on [insert topic, e.g., ‘quantum
computing policy’ or ‘feminist literary theory’]. I’d like to:
1.

Summarise the key arguments or findings in clear, accessible language suitable for a
non-specialist audience.

2.

Convert that summary into a podcast script (around 3–5 minutes) that could be used
for educational outreach or student engagement.

3.

Include a short introduction and closing statement for the podcast, and suggest a
tone (e.g., conversational, formal-academic, journalistic) appropriate for the
audience.

The goal is to develop both understanding and communication skills by translating
complex ideas into multimodal formats.

Case study summary: Teaching – Enhancing Engagement through Streaming Avatars
and Synchronous AI Assistants in Online Education by Dr Ioannis Glinavos, University of
Westminster
This case study explores the use of streaming avatars and synchronous AI assistants to
enhance learner engagement in online legal education. Dynamic avatars and AI tools
were created to facilitate an interactive and supportive learning environment that offer
real-time assistance, enhance social presence, and reduce the sense of isolation
common in online study. By incorporating multimodal elements such as image, sound,
space, speech, text, and video, the approach aims to foster a greater sense of community
and personalisation. The use of avatars also helps mitigate presentation fatigue and
privacy concerns. This strategy offers an adaptable, inclusive model that supports learner
satisfaction and well-being. Educators can adopt this approach by selecting appropriate
platforms, developing context-sensitive prompts, and iteratively refining the experience
based on learner feedback.
(Full case study is included in Appendix A.)

25


Case study summary: AI Visual Metaphor Activity by Elora Marston and Jonathan
Rhodes, University of Wolverhampton – Teacher Education
This activity invites participants on the SEDA Learning to Tutor Online course to create an
AI-generated visual metaphor that captures an aspect of their experience as online
learners—such as a key emotion or challenge. Using tools like Adobe Firefly, participants
generate an original image and post it to a shared discussion space with a short, written
explanation and a suggested application for their own teaching. Tutors model the task by
sharing their own visual metaphor and actively engage in the discussion. This multimodal,
formative activity promotes critical reflection, creativity, and empathy, supporting
participants to develop socially constructivist and compassionate online teaching
practices. The focus on generating rather than sourcing images encourages deeper
personal engagement and connects AI use to meaningful learning design.
(Full case study is included in Appendix A.)

4.2 Why do it? Benefits and opportunities of multimodal
teaching and GenAI
Integrating multimodal teaching strategies and GenAI into higher education presents
opportunities for enhancing learning and fostering creativity. We identified specific
themes from our data-gathering (survey, focus groups, case-study collection). Our
sample of educators identified the following range of compelling benefits, both
pedagogical and practical.

4.2.1 Enhancing pedagogy through new diverse modalities
Multimodal teaching can expand pedagogical possibilities by moving beyond traditional,
text-centric formats (Hall, 2024). Introducing varied media, such as video, podcasts,
infographics, interactive simulations, and performance-based tasks, can make learning
more inclusive and engaging (Yun et al., 2025). These modes support different learning
preferences and may foster deeper, more authentic engagement with content (Prinsloo et
al., 2024). As participants in our focus groups noted, students responded positively to
tasks that felt ‘real-world,’ such as building websites or designing exhibitions, which also
better prepare them for future professional contexts. GenAI can amplify these possibilities
by enabling faster content transposition e.g., summarising texts into visual formats or
adapting materials into different media, which allows educators to more easily offer
diverse entry points for learning. It also offers opportunities for showing any imagined
things that were not previously possible.

26


4.2.2 Lowering barriers to multimodal creation
While utilising multimodal learning resources can be exciting, they often require skills in
design, media production, or communication that educators or students may lack
(Walter, 2024). GenAI can help level this playing field. For instance, educators who might
have little experience with graphic design or visualisation can use GenAI tools to generate
imagery, reducing reliance on costly software or specialist knowledge. Similarly, educators
can use AI to prototype materials or experiment with new formats without needing
advanced production skills (Koppam et al., 2024). This allows a wider range of educators
to participate confidently in rich, multimodal tasks.

4.2.3 Boosting creativity and idea generation
One of the most consistently praised aspects of GenAI across our data-gathering was its
role as a creative partner. Whether brainstorming potential user personas for a design
brief or offering variations of a visual identity, AI can prompt new directions that may not
have occurred otherwise (Bolick & da Silva, 2024). This ideation support was seen
especially valuable in early stages of student projects, encouraging divergent thinking
and experimentation. Educators also found GenAI helpful in breaking out of habitual
patterns – generating fresh and different perspectives or formats that inspired them to try
new pedagogical approaches.

4.2.4 Increasing efficiency to enable human-centred teaching
Time is an endless constraint for educators. GenAI can offer tangible efficiencies in
planning, content creation, and feedback preparation (Pinski & Benlian, 2024). Participants
described how AI helped them generate quiz questions, reformat content for different
media, or produce draft scripts and case studies. These efficiencies can potentially free up
educator time for more personalised interactions, such as mentoring, formative coaching,
or supporting struggling students. In this way, GenAI can create space for what one
participant called ‘relationship-rich pedagogy’, where educators invest more in the
human dimensions of teaching and learning (Kim et al., 2025).

4.2.5 Enhancing student interest and engagement
Multimodal tasks, especially those involving creative output, often sparked student
interest and engagement. From collaborative use of Post-It notes and diagrams to
science students designing exhibitions, educators observed that alternative formats
energised classroom participation (Yan et al., 2024). While not all students immediately
embrace novel media, scaffolding, choice, and visibility of value helped overcome
hesitation. In addition, when students are supported in using GenAI to enhance their
outputs (e.g., improving visuals or experimenting with tone), they often gain a sense of
empowerment and ownership over their work (Hsiao & Zhang, 2023). Equally, our role as

27


educators lies in engaging students in the wider discourse on GenAI use within disciplinary
contexts. This requires developing students’ agency and evaluative judgement of the
utility of GenAI within specific contexts.

4.2.6 Preparing students for a digital and AI-augmented future
By incorporating Generative AI into multimodal learning, educators are better positioned
to equip students with crucial skills for the future. These include prompt engineering,
critical evaluation of AI outputs, ethical decision-making, and reflective use of digital tools
(Tan et al., 2025). In professions where creativity, adaptability, and communication are
increasingly mediated by AI, familiarity with these tools is a valuable and relevant skill.
Exposure within educational settings allows students to explore and question AI’s role in a
safe, guided environment.
Alongside these potential benefits, we recognise that there are naturally some challenges
that also come with using this approach. We have collected these together in a dedicated
part of this guide; please see Section 7.1 for further challenges of using GenAI in
multimodal teaching, learning, and assessment & feedback.

4.3 How to do it? Practical tips for educators using GenAI in
multimodal teaching
Figure 3 suggests five key practical areas for educators utilising Generative AI in
multimodal teaching. It offers ‘Time-Saving Tips’ with strategies for streamlining teaching
processes using AI tools and explores ‘Human-AI Collaboration’, providing tips for effective
teamwork between humans and AI. The infographic also delves into ‘Mirror Cases’,
showcasing examples of AI use by teachers for student demonstration, and provides an
‘Evaluation Checklist’ with criteria for assessing AI’s impact on education. Finally, it
addresses ‘AI Literacy Teaching’, outlining methods to enhance students’ understanding
and critical thinking about AI.

28


Figure 3. Five key practical areas to consider for educators utilising Generative AI in multimodal
teaching.

4.3.1 Save time with smart use of AI tools
Educators may be able to free up valuable time by using GenAI to streamline resource
creation and administrative tasks. Your institution potentially offers access to general
tools such as Microsoft Copilot or Google Gemini, which can be used to generate training
blurbs and lesson plans. Need visuals? Use text-to-image GenAI applications to create
copyright-free images, or build slides quickly. For more engaging content, there may be
opportunities to create short explainer videos using animated characters. If you’re
planning a lecture, other systems can help you build curriculum content or assessment
materials efficiently. These tools may help you shift your energy from repetitive tasks to
more impactful teaching activities.

29


Tip: If you’re short on time, GenAI software can potentially convert existing documents into
study aids or activities, such as FAQs, quizzes, or even dedicated study guides
automatically.

4.3.2 Mirror AI use: Show your process
Model transparency by showing students how you use GenAI tools in your own workflow.
Whether you are drafting questions or designing graphics, narrate your steps and
decision-making process. Invite students to do the same—ask them to reflect on how they
used GenAI in their assignments and why.
Tip: Use class time to compare AI- vs human-created content. This sparks rich discussion
and teaches students to evaluate outputs critically.

4.3.3 Use checklists to guide ethical and sustainable AI use
Before sharing or submitting AI-generated materials, ask yourself:
•

Does this content include sensitive or copyrighted information?

•

Could this image or text unintentionally reinforce a stereotype?

•

Is this the best use of AI, or is it just the most convenient?

Avoid feeding personal data into public GenAI tools like ChatGPT. When creating images,
especially of people or identities, use platforms that allow more ethical control. And
consider sustainability— avoid elaborate or purely decorative image generations - these
can have a surprising carbon cost.
Tip: Develop a simple checklist for yourself or your team covering ethics, inclusion, factual
accuracy, and relevance. Use it every time you generate content.

4.3.4 Design for Human-AI collaboration, not substitution
Think of AI as a co-pilot, not a chauffeur. Encourage students to engage with their work
multimodally by using GenAI tools. For instance, using tools to process and present
information in diverse formats. A student working on a research project could upload their
notes and then generate a mind-map to visualise key themes, create a report to
summarise their findings, or even produce a short video to present their project’s
conclusions.
For interview preparation, guide them to go beyond text: after inputting a job description
into ChatGPT to generate potential questions, they can use the GenAI’s voice feature to
rehearse their responses out loud. This interactive, audio-based practice simulates a real
conversation, allowing them to refine not only their content but also their delivery and
tone.

30


Tip: Frame GenAI as a ‘thinking partner’. Make it clear that students must do the
intellectual heavy lifting – they can use GenAI to get help getting from A to B, but not to
skip from A to Z.

4.3.5 Teach AI literacy as a core skill
Don’t assume students (or colleagues) know how to use AI effectively or responsibly. Build
in activities to practice prompt engineering – the clearer the input, the better the result.
Highlight the strengths and limitations of different tools. For example, you can show how
some tools provide sources, or demonstrate how AI can ‘hallucinate’ facts, especially in
images.
Tip: Incorporate tool-specific training into assignments. For example, when asking
students to build digital field notebooks, teach the platform first and check for
accessibility needs.
Bonus Tip: Ask students to document how and why they used AI in their work. This builds
reflection and reinforces responsible habits.

31


5. Multimodal Generative AI in Learning
Multimodality in higher education that explicitly focuses on student learning can consist of
formats, such as students encountering, exploring, critically evaluating, creating and
expressing their learning through multimodal artefacts (Lim & Tan-Chia, 2023; Lacković &
Olteanu, 2023). This could include students being asked to bring a multimodal artefact to
the classroom for discussion, or a critical analysis of a new concept. For example, see
Lacković’s (2020) inquiry graphics method for analytical and critical thinking about
students’ digital images with their conceptual interpretations, and Varga-Atkins and
O’Brien’s (2009) work in ‘elicitation techniques’ in research that are applicable in learning
- eliciting reflection through multimodal artefacts, such as diagrams. Students can also
be encouraged to evaluate and compare information provided on the Internet and
various GenAI platforms, such as discussing a published blog post or GenAI outputs with
peers and teachers.
For multimodal learning to be realised in this case, it is important that students express
their learning by creating a multimodal artefact (e.g. a poster, a multimedia presentation,
an inquiry graphic or an infographic) or else experience that artefact multimodally. To
illustrate, students could be undertaking their learning in multimodal environments,
whether it is a physical classroom with the lecturers and peers drawing on the modalities
of spatial arrangements, movement, gestures for communicating, or virtual classroom
utilising semiotic technologies (technologies and platforms with multimodal capabilities)
or immersive learning environments (augmented or virtual reality). GenAI can potentially
make meeting this necessity much easier; there are ample opportunities to engage
students in multimodal learning designs with current technologies, and particularly with
GenAI platforms. However, as we have argued, these technologies are not to be embraced
at face value and adopted uncritically.
There are numerous reasons why educators and educational developers need to embed
multimodal learning designs in their practice. The rapid development of multimodal
technologies creates an ever-evolving technological landscape that engages our senses
through visual, audio, tactile and other modalities in addition to linguistic communication.
Most students actively use highly audiovisual platforms for communication and
entertainment on daily basis, such as YouTube, Instagram, Snapchat, Netflix, WhatsApp,
etc. in Western contexts, and platforms such as Xiaohongshu, Bilibili and WeChat in China,
and a plethora of GenAI platforms globally. Through their usage, students are
overwhelmingly positioned as – and are occasionally deliberately forced into becoming –
fast consumers of information, without many opportunities to pause, create and reflect on
digitally created and co-created information and artefacts. Finding time to create and
reflect is particularly important with regards to students’ creative engagement with
technologies and their critical thinking skills and the nurturing of criticality in general.
32


Furthermore, a growing body of research evidences the value of grounding what are
commonly termed ‘abstract’ concepts across disciplines to sensory experiences that are
best expressed through multimodal artefacts. As we have emphasised earlier, this means
engaging students to create such artefacts to discuss and evaluate their own learning
with peers and teachers. For example, authors have provided empirical evidence and
case studies showing how diverse media can be used with image-salient platforms and
artefacts (see Lacković (2020) for embedding digital images in learning design that
evidence both the value and challenges of such a design). Lacković and Olteanu (2023)
provide case studies on various multimodal artefacts for higher education learning as
applied by diverse teachers in Part 3 of their book, while Tzirides et al. (2024) suggest a
coupling of educational concepts with GenAI images for student reflection, akin to the
creation of inquiry graphics with GenAI.
The point of embedding multimodal designs is not simply to use any digital platform and
then evaluate it, as these kinds of studies may obscure the fact that technology
implementation depends on good pedagogic design. The importance is placed on how
technology is implemented and how technology affordances are used by students to
benefit student learning and in what way.
All in all, pedagogy should lead technology (rather than the other way round), as pointed
out by many scholars. One of the main aims of pedagogical designs is to go beyond
functional, management and efficiency uses of technology in higher education. As Selwyn,
Henderson and Aston (2017) emphasise in their investigation of student technology uses,
students may still not be using varied technologies for creative and reflective tasks but
more for management and assistance tasks, which are functional rather than creative
and innovative uses. Certainly, there is merit in functional uses, and there is some
innovation there too, but higher-order thinking skills such as creative and critical
engagement are at the forefront of desirable student growth in higher education.
The rise of GenAI tools that offer a wide range of visual, audio, linguistic and hybrid outputs
has ushered a new era of opportunities and challenges for multimodal learning designs.
For some students, GenAI can lower technical barriers in multimodal tasks (e.g., designing
infographics or generating voiceovers), enabling them to focus on ideas and meaningmaking rather than format limitations. In the section that follows, we will report on
examples from our research study on how GenAI can be utilised in these multimodal
learning processes that can support personalised, adaptive and critical learning.

5.1 What is it? GenAI in multimodal learning
These ideas for multimodal uses of GenAI for learning are written in a way that either
students can initiate or undertake these activities or that educators can design these
activities for students to undertake.

33


5.1.1 Personalised learning and self-paced study
GenAI could be used for personalised learning and self-paced study, such as for revision
or goal setting. This could mean using GenAI tools to help students create schedules and
‘to dos’, and the advantage is that whatever the format of the original material, digital or
handwritten text or visuals, GenAI can quickly convert one format to another, and is adept
at re-arranging the materials in a newly required format (table, visual, timeline etc.), or
according to a new organising principle. Examples here also include students (or
educators) creating flashcards and quizzes to help self-assess students of their learning
(e.g., at time of writing, NeuralConsult is an example of specifically aimed at medical
students), or creating explainer videos using GenAI platforms (e.g. Heygen for video
characters). Other examples include educators creating personalised podcasts with help
of GAI for each learner depending on their subject area or level of knowledge.

5.1.2 Just-in-time support and study help
GenAI can also be used for just-in-time support, explanations and study help as a
resource that is accessible to students 24/7 to break down complex ideas, conceptual
questions and explain them according to the level of current understanding of the
students. GenAI can also offer explanations by applying and demonstrating it to contexts
the student would understand or demonstrate concepts in different ways that scaffold
student learning. As above, from a multimodal perspective, GenAI can be ‘fed’ multimodal
artefacts such as diagrams, maps, concept maps, infographics, equations, and
calculations to produce explanations, and vice versa, the produced explanations can be
asked in different formats (e.g. ‘create an image to help me visualise X/concept’) (that
said, the outputs may depend on the multimodal affordances of the GenAI platform used
and the underlying model).
An advantage of using GenAI for explanations is that it can be easy to prompt GenAI to
create or mock up some data if, for instance, the concept would benefit from generating
worked examples, or take students through learning sequences and/or supported guided
practice.
Sample GenAI Prompt:
I’m struggling to understand what a “normal distribution” is in statistics. I kind of get that
it’s a bell curve, but I don’t really understand what it means in real life. Can you explain it
simply, like I’m a first-year nursing student? And could you also show me a diagram or
make up some data to help me see how it works in a real example?

34


Case study summary: Learning – Using Generative AI to Aid Studying and Learning for
Medical Students by Ayaat Jaway, University of Liverpool – Medicine
This case study explores how medical students can use generative AI, specifically
ChatGPT, to support students’ revision and understanding of complex topics. The
approach utilises text and image generation to make studying more comprehensive and
personalised. Students can input their own notes or topic keywords to generate simplified
explanations, tailored diagrams, and exam-style questions, supporting a process of
scaffolded learning that builds in complexity over time and helps enhance
comprehension, engagement and retention. The key benefit of this method lies in its
adaptability: by using prompts, learners can direct the GenAI platform to explain topics at
varying depths, visualise processes and practise with AI-generated exam questions.
Although the case study is in medical education, the method is easily transferable to other
disciplines where complex, content-heavy material presents a barrier to learning.

(Full case study is included in Appendix A.)

5.1.3 Critical thinking, metacognition and reflection
GenAI can also be used for prompting critical thinking and metacognition, prompting
deeper reflection and learning. This may also take the form of challenging students’
current assumptions or helping them consider multiple perspectives and take them
through structured reasoning. GenAI can be used, for instance, to offer up different user
types, scenarios or requirements so students could build on these different perspectives in
35


their reflections. The advantage of GenAI here can also be that these requirements can be
dynamically acted out in a conversation and dialogue, so students can adapt, deepen
and re-evaluate their views, discover flaws and connections in their logic, and so, learn as
they go along. Conversely, students could be critically reflecting on GenAI outputs,
whether fact-checking, checking for bias, inclusivity or representation, developing critical
AI literacies.
Screenshot:
In the example below, visual ‘character cards’ are used (e.g., NGO worker, climate activist,
government official) each with an embedded voice prompt or soundbite generated by AI.
Students use these as stimulus to reflect or role-play with GenAI as part of a discussion on
climate adaptation (or whatever the topic may be). This could add a sensory dimension
to AI-driven perspective-taking, prompting emotional, cognitive and critical engagement.

Dialogue excerpt:
GenAI Climate Activist:
‘Climate adaptation needs to be rooted in justice. How can we protect the most
vulnerable communities if we keep prioritising economic growth over resilience?’
Student:
‘That’s a good point. But what if the government says funding is limited and some
compromises are necessary?’
GenAI Climate Activist:
‘Then we need to question who those compromises are affecting. Are we really adapting,
or just delaying the inevitable for those already at risk?’

36


5.1.4 Collaborative and dialogic learning
Another use of GenAI in multimodal contexts is for collaborative and dialogic learning,
such as enacting a peer conversation for learning, testing ideas or receiving immediate
feedback that draw on multimodal resources and communication to support knowledge
construction. Examples include conducting a dialogue using speech with AI to act as a
patient with particular illnesses or any kind of similar Question & Answer format that can
ask GenAI to adopt certain roles in different contexts. This technique can also be used by
learners to prepare for a particular exam with questions that contain multimodal formats,
rehearsing answers from past exam papers in ways that include either multimodal input
or output (aural or visual e.g. diagrammatic questions or reasoning).
Case study summary: Learning – Live Storytelling through Generative AI-powered
Avatars by Simon Campion, Radoslaw Dorociak & Georg Meyer, Virtual Engineering
Centre, University of Liverpool – Engineering

This case study showcases a multimodal learning experience where learners interact with
a digitally recreated historical character, Mary Seacole. Using Epic MetaHuman
technologies, the team developed a visually rich, AI-powered avatar capable of text and
speech interaction in 40 languages. The system draws on preloaded historical content
and an adaptive Q&A database, enabling learners to ask questions naturally and receive
contextually appropriate responses. The approach combines video, image, speech, and
text to support experiential, accessible, and personalised learning. Originally developed for
a museum setting, the method can be applied across disciplines to bring subject content
to life through virtual dialogue and immersive storytelling.
(Full case study is included in Appendix A.)

37


5.1.5 Creating modally accessible and inclusive resources
A great benefit of using GenAI platforms is that it can support students with making
learning materials more modally accessible and inclusive, for instance through
generating alternative formats, translations, or sensory-accessible resources, for instance
by adapting the format, level, language or delivery mode. Students can translate
materials into their own/another language to help understanding. Live transcription can
supplement lectures; transcription and GenAI-created summaries of supervisory
meetings or tutorials can support the learning process. This can be audio-to-text but with
GenAI content can be quickly reformatted and re-organised based on some additional
requirement or perspective. Students with dyslexia, but also any student in the spirit of
Universal Design for Learning (UDL), can benefit from recreating learning materials in
different formats (e.g. slides into text/audio) (Rusconi & Squillaci, 2023).

5.1.6 Creating, exploring and transposing study content in different
modalities
GenAI could be used for creating, exploring, understanding and engaging with study
content that works with or transposes content from/into different modalities e.g.
summarise, expand, clarify topics from/into diagrams, images, videos or speech/audio.
An example of this would be summarising articles or presentations/slides into a podcast
(e.g. using Notebook LM), or vice versa, creating a slide or presentation from an existing
text or lecture notes (e.g. Gamma), or creating multimedia, websites, designs. Other
examples include students documenting their learning process in a visual/comic strip
style using prompts or using AI voice to narrate their own presentation if they are less
confident in speaking. There are examples of AI-functions within existing platforms that
students can use for learning, e.g. PowerPoint presentations/MS Teams that offer
personalised, immediate feedback to students such as a public-speaking coach to help
students practise presentations with immediate feedback (Rehearse Coach) or reading
(Reading Coach).

38


Screenshot:
Example of taking a set of slides (e.g. on the Nominal Focus Group technique) and
creating an audio overview or podcast using NotebookLM (or similar GenAI platform) to
engage with learning in different modes and formats

Another format could be mindmaps, so learners could prompt the GenAI platform to turn
the slides or content into a mindmap - see below.

39


Case study summary: Learning – Creating inquiry graphics with GenAI: learning with
images by Maede Seyedeh Mirsonbol, University of Torino, and Nataša Lacković, Lancaster
University
This case study showcases a multimodal learning activity where learners created (cocreated) images using GenAI. The purpose was for learners to share their own
interpretations of one aspect or characteristic of the concept through an image that they
imagined and then instructed GenAI to create. Learners then shared their images and
interpretations (narratives) about how the image represents something about the
concept with other learners.
Learners can apply different activity designs, such as a ‘gallery talk’ where all the
generated images are shared with the whole group or within groups, including the
teacher’s interpretations/input. In this particular case, the concepts tackled were: UN
Sustainable Development Goals (SDs), inclusivity, and developing an inclusive classroom.
Originally developed as part of a pedagogic activity termed ‘inquiry graphics pedagogy’
or ‘image-based concept inquiry’ (Lacković, 2020), this method can be applied across
disciplines to surface learners’ interpretations, prior knowledge and experiences, linked to
key educational (programme/course) concepts.

(Full case study is included in Appendix A.)

5.1.7 Advice provision – GenAI as assistant
Another way students reported using GenAI in our focus group exercises was seeking
advice in the case of interpersonal conflicts or study-related challenges from a neutral
viewpoint and in confidence, without any other persons involved. When conducting our
research, a number of students reported utilising GenAI Chatbots as a counsellor or
40


advisor in their given life situation, or even for social, personal benefits such as creating
shopping lists, helping with recipes or seeking advice in friendship or relationship issues or
when in conflict with other people. Such uses can also apply in a study context. For
instance, when students come up against conflict during group exercises, they may want
to seek advice on next steps or strategies that may work in the given situation and with
particular characteristics, as the AI might act as an impartial arbiter in a conflict scenario
– although students should always remain critical of GenAI outputs.

5.1.8 Emotional support for study
GenAI could be used for emotional study support, such as helping students stay
motivated and engaged, for instance by eliciting encouragement or providing a
confidence boost. Students could ask for tailored responses on educational topics that
suit their mood or ask for daily reminders. Or, when students get negative feedback or a
low mark on an essay (or when PGR students may get rejected by reviewers after
submitting manuscripts), they could use GenAI to break down the feedback into
actionable pieces and a table, which they can revisit when feeling emotionally ready to
tackle feedback. This could also be used in the pre-submission context, where students
get an AI to examine their draft work ahead of sending it in to ensure that they have met
the required criteria, which may act as a confidence boost.

5.1.9 Help with organisation and admin tasks
GenAI can perform well with supporting students with organisational and administrative
tasks that support their learning. For instance, repetitive tasks such as formatting
references, documents or making schedules, checklists, suggest a timeline or actions, or
breaking down larger pieces of learning into component parts and smaller tasks, and
present it effectively in a given format, perhaps highlighting critical/important tasks and
working towards given assessment or task deadlines. These outputs then can be asked in
multimodal formats, such as timelines, maps, or diagrams.
Sample GenAI Prompt:
Create a visual timeline for producing this guide by the 31 August 2025 deadline, starting
today (10 June 2025) including tasks such as drafting, internal review, student advisor
review, revisions, find edits and accessibility checks, design work.

41


5.1.10 Specialised, disciplinary support for learning
Finally, there are disciplinary use cases for multimodal learning with AI where
LLMs/chatbots have been used - see two examples below from marketing and
architecture. In some cases, GenAI platforms have been custom-built and pre-trained to
suit disciplinary epistemic practices and artefacts. Hsiao and Zhang (2023) for instance,
customised models in architecture so that students develop a deeper understanding of
design aesthetics and enhanced their design thinking abilities. These examples also show
the links to employability skills.
Case study summary: Learning – Enhancing Learning with AI Tools for Idea Generation
by Katherine Geer, et al., Liverpool John Moores University – Marketing
This case study explores how AI tools were integrated into a design thinking framework to
support multimodal learning in a ‘Media Production Management’ module. Across three
stages - Understand, Explore, and Materialise – 198 students used Generative AI tools.
including ChatGPT, MS Copilot, and NotebookLM, to develop and refine digital content
ideas for real-world clients. Students engaged with text prompts to research inclusive
campaign ideas, generated visual storyboards to communicate concepts, and created
audio narrations to test podcast content. These multimodal interactions (text, image, and
sound) supported creativity, critical thinking, and collaborative learning. Importantly,
students were encouraged to critically evaluate AI outputs, building essential AI literacy.
This approach not only enhanced idea generation but also equipped students with
relevant workplace skills, offering a practical model for using AI in content-rich, creative
disciplines. The case highlights how multimodal AI supports deeper engagement with
both process and output.
(Full case study is included in Appendix A.)

42


Case study summary: Learning – The Role of GenAI in Design for the Disassembly of
Emergency Shelters, by Farhondeh Vahdati, University of Liverpool – Architecture
This case study explores how multimodal GenAI supports architectural education and
humanitarian training by generating visual step-by-step disassembly guides from
annotated shelter diagrams. Using tools like GPT-4 with vision, students upload structural
sketches of vernacular-inspired shelters and receive AI-generated visuals that illustrate
disassembly sequences, joinery, and material hierarchy. This enhances students’
understanding of spatial relationships, modularity, and design for disassembly (DfD) –
concepts often difficult to grasp through text alone.
The approach aligns with constructivist learning, promoting active exploration and
iteration. It also has real-world applications: simplified, language-free guides could
support non-specialists in emergency contexts. This demonstrates how AI can be a cocreative learning partner, transforming complex 3D concepts into accessible 2D visuals.
Multimodal interaction (image + text + interaction) makes learning more inclusive and
adaptive, offering a scalable model for integrating GenAI in both academic and disasterresponse settings.
(Full case study is included in Appendix A.)

5.2 Why do it? Opportunities and Benefits of GenAI in Multimodal
Learning
GenAI offers a range of benefits that align with multimodal learning. Drawing on recent
research and insights from our project, the key advantages and challenges are included
below.

5.2.1 Lowers technical barriers
•

GenAI can enable students and educators to produce multimodal artefacts (e.g.
videos, infographics, scripts) without needing advanced technical or media
production skills, reducing reliance on specialist tools or support (Lee & Li, 2024).

•

This opens up multimodal learning for a wider range of students, especially those
who may have previously lacked the confidence or tools to express ideas beyond
text.

5.2.2 Enhances content production efficiency and scalability
•

GenAI can rapidly convert content across modalities (text to image, notes to audio,
slides to summaries), allowing students to focus more on conceptual
understanding than formatting or transcription tasks (Bolick & Da Silva, 2023).

43


•

This makes multimodal learning more time-efficient and scalable for both
students and educators, particularly in resource-constrained settings that have
GenAI access.

5.2.3 Supports personalised, cross-media and adaptive learning
•

Through dialogic interaction, GenAI can tailor content explanations, examples, and
resources to suit a learner’s knowledge level or preferred modality of expression
(Alasadi & Biaz, 2024; Liu et al., 2024).

•

Students can also revisit and reshape AI-generated responses in different formats,
enabling adaptive and self-paced learning suiting their personal preferences.

5.2.4 Fosters reflective and critical thinking
•

Multimodal GenAI activities have been shown to prompt deep reflection and
metacognition (Milesi et al., 2024). Students can consider how best to express and
structure their ideas across different modalities and formats.

•

Critically engaging with AI outputs also helps build AI literacy, as students learn to
prompt, evaluate, and refine outputs (Tan et al., 2024).

5.2.5 Promotes inclusion and neurodiversity
•

GenAI can provide alternative formats that reduce cognitive load, particularly for
neurodivergent learners or those with sensory preferences (Jafarian & Kramer,
2024); for example, by converting dense reading into audio summaries or visual
maps.

5.2.6 Deepens disciplinary understanding
•

In subject-specific contexts, such as design education, the ability to quickly
generate and iterate on visual artefacts supports students in developing
disciplinary thinking (e.g. aesthetic reasoning, visual storytelling) (Hsiao & Zhang,
2023).

5.2.7 Encourages experimentation and practice in a low-stakes
environment
•

GenAI offers a non-judgemental space for exploration, where students can test
ideas, take creative risks, or pose ‘silly’ questions without fear of embarrassment—
valuable for building confidence and creativity.

5.2.8 Can support wellbeing and social situations
•

Some students report using GenAI for emotional and motivational support, such as
seeking encouragement, breaking down feedback into manageable parts, or
asking for help when overwhelmed or in problematic situations. While not a
44


replacement for human support, GenAI can supplement wellbeing or conflict
resolution strategies in educational contexts 24/7.
Alongside these potential benefits, we recognise that there are naturally some challenges
that also come with using this approach. We have collected these together in a dedicated
part of this guide; please see Section 7.1 for further challenges of using GenAI in
multimodal teaching, learning, and assessment & feedback.

5.3 How to do it? Practical tips for educators and students using
GenAI in multimodal learning
These ideas have been written either for you as educators to embed into your curricula, or
for your students to try directly.

Figure 4. A model for a cycle of multimodal learning using Generative AI, starting with clarification of
the learning goal and leading through selection, experimentation, evaluation, expression and ending
with reflection.

45


5.3.1 Clarify your learning goal (before using GenAI)
What to Do: Ask yourself: ‘What am I trying to learn or communicate?
Am I trying to acquire knowledge? Investigate or research? Collaborate? Produce
something? Practice something? Discuss or evaluate my learning?
Our Suggestion:
•

Learners: Write down your goal in one sentence. Example: ‘I need to understand
how photosynthesis works for my biology exam—maybe a diagram would help.’

•

Teachers: Have students articulate their goals before using AI. Example: ‘In one
sentence, explain what you want AI to help you with.’

5.3.2 Choose the purpose of using AI for your learning goal
What to Do: ‘Am I trying to create a resource or representation with AI? Am I using AI to
teach me something? Am I going to express my learning or do some action?’ Also
consider which tasks will GenAI be better than humans, and which tasks should remain
better performed by humans.

5.3.3 Choose the right tool and format
What to Do: Pick an AI tool whose multimodal affordances match your goal. Need an
animation? Try HeyGen or Canva AI. Need flashcards? Use Quizlet’s AI. Would audio or
video help you with this goal better?
Our Suggestion:
•

Learners: Match the tool to the task. Example: ‘I’ll use ChatGPT for a summary and
DALL-E to draw the process.’

•

Teachers: Recommend tools for specific tasks. Example: ‘Use Suno AI to turn poetry
into songs for memorisation.’

Note: consult your institutional policy and guidance and check you are not submitting
private data (see Sections 2 and 7 for more details)

5.3.4 Experiment with prompts - don’t settle for the first output
What to Do: Treat prompts like a conversation. Start broad, then refine. Example:
‘Explain supply and demand’ > ‘Now add a real-world example’ > ‘Turn this into a table
comparing two countries.’

46


Our Suggestion:
•

Learners: Use the ‘Purpose Hack’: Add ‘for exam revision’ or *’for a 10-year-old’* to
tailor outputs.

•

Teachers: Assign prompt-iteration exercises. Example: ‘Improve this AI output in
two steps.’

5.3.5 Critically evaluate AI outputs
What to Do: Ask: ‘Is this accurate? Does it match what I already know? Does the format
help?’, ‘Have I met my learning goal?’, ‘What discourse underpins this Large Learning
Model?’
Our Suggestion:
•

Learners: Cross-check AI with trusted sources. Example: ‘This AI timeline of WWII
skips a key event – I’ll add it.’

•

Teachers: Create ‘AI Fact-Check’ assignments. Example: ‘Find one error in this AIgenerated summary.’

5.3.6 Express what you have learned in a new format
What to Do: Use AI to transform your knowledge. Example: Turn notes into a podcast script,
comic strip, or quiz. Expressing in multiple formats (e.g., explaining aloud + drawing) can
help deepen your learning.
Our Suggestion:
•

Learners: Try the ‘Switch It Up’ challenge. Example: ‘I studied this via text—now I’ll
explain it to my dog using only doodles.’

•

Teachers: Assign multimodal projects. Example: ‘Explain the water cycle using AI to
generate both a paragraph and a mind-map.’

5.3.7 Reflect on the process
What to Do: Ask: ‘Did AI save time? Where did I still struggle? What format worked best?
How would I do it differently next time?’
Our Suggestion:
•

Learners: Keep a ‘GenAI Learning Journal’. Example: ‘ChatGPT summaries saved
time, but I had to redraw the graphs myself.’

•

Teachers: End lessons with a 5-minute reflection. Example: ‘Share one way AI
helped and one way it didn’t today.’

47


6. Multimodal Generative AI in Assessment and
Feedback
Multimodality in the assessment space can take a number of different forms. Naturally,
assessment artefacts themselves can be multimodal, by asking students to create
something that combines multiple semiotic forms to produce a greater meaning than if
the artefact was limited to a single mode (Kress & Van Leeuwen, 2001). These artefacts
might include any combination of text, image, sound, colour, movement, video, etc.,
arranged so that they either complement or reinforce one another, in order to produce
recognisable named artefacts like posters, infographics, videos, blogs/vlogs, podcasts,
portfolios, websites, social media posts or applications, among others.
But it is not only the assessment artefact that can be multimodal. Multimodal
assessments can also include elements of multimodality in their completion, by asking
students to work with multimodal resources or collaboratively work with each other over
different media to ensure that the assessment is completed successfully. Live
assessments, i.e. where students perform or present something in a live situation, can also
be multimodal, where students must combine images often presented on a slide deck
with speech and gesture to help the viewer construct the most meaning out of both forms
of information.
While feedback in higher education still very much relies on text, several examples listed
below demonstrate the impact and potential of multimodal feedback. This often involves
feedback that combines text with audio and/or video, where teaching staff record
themselves discussing the students’ submission alongside the submission itself.
Audio/video feedback is already in use in some contexts, but it is not particularly common
– perhaps due to nerves, discomfort, or the more practical necessity to circulate examiner
feedback to external moderators/examiners – often straightforward with written-text
assignments, but can be challenging with multimodal feedback. In other contexts, where
long-form feedback is not possible (for example, on multiple-choice question (MCQ)
assessments) multimodal feedback might constitute smaller or abstract visuals, sounds
or haptic vibrations to indicate when a student has got an answer correct or incorrect.
These can add to the richness of feedback and go some way to provide an alternative to
in-person dialogic feedback, which is not always possible to provide due to high student
numbers (Martin, 2020).
Both multimodal assessment and multimodal feedback requires students to develop
multimodal literacy – their capacity to understand the way that modes interact with each
other when producing an assessment artefact, as well as the information that those
modes convey within the context of their subject discipline (Ross et al., 2020).

48


As with both Teaching and Learning, Generative AI has the potential to affect the way
multimodal Assessment and Feedback is conducted in the higher education space,
particularly in designing multimodal assessments, completing multimodal assessments,
and reflecting on multimodal assessments. In essence, GenAI can help develop
academics’ capacity to design more innovative, more authentic and more relevant
multimodal assessments, and can also help students to complete assessments more
efficiently and open up new avenues for students to access different semiotic modes
when completing their assessments.

6.1 What is it? GenAI in multimodal assessment and feedback
6.1.1 Generative AI in Multimodal Assessment
Generative AI has hitherto been largely perceived as a threat to the integrity of summative
assessments, particularly as it is both difficult to detect its illicit use in creating
assessment artefacts, and because redesigning assessments into forms that are
simultaneously perceived to be ‘authentic’ and that also create contexts where GenAI
simply cannot be used has proven elusive (Kofinas et al., 2025). However, GenAI has
substantial potential to enhance multimodal forms of assessment in a number of
contexts, and multimodal assessment in of itself has the potential to create assessments
where GenAI can be permitted with confidence that students cannot, and will not, attempt
to use it illicitly.
At the most basic level, GenAI might simply help the student who is completing the
assessment with creating and enhancing different aspects of the overall final multimodal
submission. Multimodal assessments often make use of semiotic forms that are difficult or
time-consuming to create from scratch, but which lend themselves well to AI-enhanced
creation, such as images, videos or audio. Hitherto, students wishing to use this form
would be required to create, record and edit such content, and the amount of time and
energy spent on creating this aspect of the overall submission may not have been
proportional to the contribution it made to the piece as a whole. In addition, such modes
of information were not always accessible to students who were less digitally literate than
some of their peers, which put them at a slight disadvantage. Supported by GenAI,
multimodal tools can empower creative expression and have the capacity, then, to open
up different modes of information to students who previously would not have considered
using them, and new forms of multimodal assessment to teaching staff who want to
move beyond the textual. It also helps to redress the imbalance between the time and
effort to create such content vs. how much it is actually worth to the overall submission,
allowing both student and teacher more time and space to focus on the content
contained within the semiotic form, not the semiotic form itself.

49


Generative AI might also help teaching staff to conceptualise multimodal assessment
ideas and help them move beyond simple text-based assessments, exams or more
traditional assessment modes that are there only by sheer virtue of being wellestablished within the pedagogic consciousness. GenAI can be particularly useful to
examine learning outcomes on modules and suggest new ideas on the most effective
way to assess them, and this is particularly effective if the AI contextualises the response
against wider levelling frameworks such as the Framework for Higher Education
Qualifications (FHEQ) in the UK, or equivalent frameworks elsewhere. GenAI can also be
particularly helpful for suggesting ways to assess outcomes multimodally, as well as
authentically, in particular disciplinary contexts.
Thirdly, Generative AI has particular value in the research or establishment phases of an
assessment artefact. Finding/organising literature for use in a final submission, for
example, is often a textual process, but GenAI can help make the process more visual,
particularly through the use of AI-enabled tools that iteratively map out and identify
connections between authors, papers, journals or concepts. Figure 5, for example, shows a
screenshot of the ‘Research Rabbit’ AI-powered literature search and organisation tool,
helping to visually represent the field in a chosen or searched topic area.

Figure 5. A screenshot of the ‘Research Rabbit’ tool, showing a map of the scholarly field within the
multimodality and Generative AI space.

Interestingly, a recent review of literature on students’ engagement with Generative AI
emphasised their predominant engagement with text-to-text models like LLMs and a
corresponding much lesser understanding of alternatives (e.g. systems that can present
textual information as sound, speech, images, etc.), despite these latter systems having
distinct advantages in developing reading comprehension, creativity and understanding
of core concepts (Heilala et al., 2025). This would imply that learning to use specifically
50


multimodal Generative AI as part of establishing an assessment artefact could enhance
both learning and assessment literacy.
Finally, GenAI might help the student access the assessment better, particularly through
multimodal means. GenAI might be used to reformulate the assessment brief, marking
criteria, or rubric into an audio form, for example, to help students to better understand
and engage with the assessment and ultimately perform better.
Generative AI, then, has the capacity to enhance multimodal assessment at every stage
of the assessment process, from conceptualisation and design, through to access and
understanding, through to completion and submission.

6.1.2 Generative AI in Multimodal Feedback
Generative AI might help reformulate simple text feedback into something multimodal, in
order for students to obtain something more meaningful. Audio feedback has often been
touted as preferable to textual feedback, given that students often find it clearer, friendlier,
and more discursive than textual feedback that simply points out errors or areas for
improvement (Voelkel & Mello, 2015). However, while it is often perceived that recording
audio feedback can save time for academics, this is not always the case - it often
produces better quality feedback, but it is not necessarily faster (Voelkel & Mello, 2015),
although audio typically generates many more comments than written feedback for the
same time commitment - we can talk much faster than we write or type.
Utilising the smartphone has seen enhanced results in the distribution of audio feedback
and efficiency timewise (Nortcliffe & Middleton, 2012). Audio has also been used for
broadcast feedback, or ‘generic feedback’, targeted at whole cohort assignment groups
(Middleton & Nortcliffe, 2010). GenAI might help with this, in that it can perhaps reformulate
feedback produced in the form the tutor prefers into other (or multiple) forms, which can
then be provided to the student.
Elsewhere, GenAI might act as an additional source of formative feedback for students
outside of their interactions with their tutor. Students might provide draft assessments to
GenAI tools that then create new artefacts based on the submission. A student might use
an AI podcast generator to listen to an AI podcast of their own work, which will provide
insights into their own argument, perspective, research context, focus and quality of
language. Interactive AIs might also be quizzed on the quality of the draft submission in
relation to a set of marking criteria also uploaded to the AI, which provides students with
an additional source of guidance on completing the assessment – although it should be
noted that this should not replace feedback from the tutor, but rather act as a
supplementary source. This also creates scope for dialogue-based feedback with AIs,
where the student is, in essence, discoursing with their own work.

51


Generative AI therefore has substantial scope to help create multimodal forms of
feedback and help to create more dialogue-based, accessible and holistic forms of
feedback outside of textual comments.

6.2 Why do it? GenAI in multimodal assessment and feedback
6.2.1 Assessment
Incorporating Generative AI into multimodal assessment practices has a number of
potential benefits:

6.2.1.1 Increased diversity in assessment practices
A multimodal assessment is more likely to have a unique form and a specific, relevant
connection to the disciplinary context in which the assessment takes place, and
Generative AI provides improved access to multimodal activities.

6.2.1.2 Increased authenticity in assessment practices
The world is multimodal, and the disciplinary contexts in which students will eventually live
their professional post-graduation lives will be too. Students are also likely to need to
engage with Generative AI in their post-university lives, too, and so using AI to enable or
enhance multimodal assessment should make it more relevant and applicable – in other
words, authentic – to students.

6.2.1.3 Improved employability and professional skills
Relatedly, using Generative AI in multimodal assessment contexts should also improve
students’ employability, as they will have engaged with both practices and technologies
that are relevant to their future professional contexts.

6.2.1.4 Increased digital literacies and capabilities
AI-enabled multimodal assessments provide opportunities for students to explore a
swathe of technologies – not just Generative AI, but more holistic software – to produce
assessments (Hung et al., 2012; Ross et al., 2020). They will need to draw connections
between technologies to create multimodal artefacts that successfully interact between
semiotic forms, thereby developing their wider digital literacies across different
technologies.

6.2.1.5 Increased student engagement
Creating more relevant, applicable, authentic and technologically-enabled (yet still
purposeful) assessments is likely to drive up students’ engagement, as they will be more
likely to see why the assessment is relevant and connected to the discipline they have
chosen to study and why it will be helpful to them in future professional or social contexts.

52


6.2.1.6 Enhanced assessment accessibility and inclusivity
There is potential to increase the accessibility of, and inclusivity of, assessments and
assessment processes, by using AI in multimodal contexts, thereby increasing students’
assessment literacy/ies. Assessment briefs, rubrics, marking criteria, etc., can all be
enhanced and adapted using Generative AI to make them more accessible, while
Generative AI can open up practices for students that might not have previously been
possible depending on their level of digital literacy and capability, or their more holistic
(dis)ability.

6.2.1.7 Enhanced assessment literacy
Assessment literacy can be enhanced through multimodal use of audio and video by the
teaching staff by providing the requirements of the assessment in alternative modes.
Generative AI podcasting tools could create a more nuanced discussion with which
students could interact.

6.2.2. Feedback
In the feedback space, Generative AI-enabled multimodality also has a number of
potential benefits:

6.2.2.1 Increased feedback literacy
Using AI to adapt feedback into multimodal forms helps students engage with it in a way
that works for them. As a result, students are more likely to make better use of, and get
more out of, their feedback if it is presented in a form with which they connect. This also
helps to make feedback more accessible, depending on the needs of the individual
student. As a result, students are more likely to engage with their feedback, as it will be in
a form they find helpful and meaningful.

6.2.2.2 Increased opportunity for dialogic feedback
There is an increased opportunity for dialogic feedback by using AI to create multimodal
feedback: students can use AI to engage with their work directly, or create artefacts about
the work to discuss with their tutor, or use AI to create an artefact to discuss with their
peers, among other activities.

6.2.2.3 Saving time in creating feedback
Using AI to help create or reformulate multimodal forms of feedback can potentially save
time for tutors who might otherwise find it difficult or time-consuming, particularly for
large groups of students or particularly complex assessments.

53


Alongside these potential benefits, we recognise that there are naturally some challenges
that also come with using this approach. We have collected these together in a dedicated
part of this guide; please see Section 7.1 for further challenges of using GenAI in
multimodal teaching, learning, and assessment & feedback.

6.3 How to do it? GenAI in Multimodal Assessment/Feedback
6.3.1 Assessment
6.3.1.1 AI for Developing Assessment Literacy
To help students understand the purpose of any assessment, a scaffolded approach to
develop assessment literacy is needed to help them make sense of assessment practices,
interpret assessment tasks and criteria, and to use feedback to improve learning. Below
are some potential ideas to try for this approach:
1.

Explain the purpose of the assessment and relation to the intended learning
outcomes. Consider using multimodal ways to do this including infographics and
screencasts.

2.

Help students need to understand the different types of assessment e.g. formative,
summative, peer, self.

3.

Help students make sense of the assessment criteria and marking rubric. Students
need to understand what is being assessed e.g. argument, structure, critical
thinking, and what this looks like at different performance levels e.g. 2.2 vs 2.1 vs 1st.
Consider how an AI generated podcast may help to explain this further.

4. Use GenAI to get feedback on your marking rubrics to clarify your expectations
more fully whilst ensuring the language used is clear and inclusive.
5. Review assessment briefs using GenAI acting as different student personas e.g.
mature student returning to study, international student whose first language is not
English, or a neurodivergent student.
6. Generate a quiz to test students’ understanding of the assessment brief and what
is required, providing an opportunity to clarify points. Alternatively, students could
use GenAI to demystify and interrogate the brief for example by asking ‘explain this
assessment in plain English’ or ‘Give me a checklist based on this marking rubric’.
Encourage students to use GenAI as a thinking partner.
7.

Provide students with information on the opportunities for formative feedback and
how to use this feedback effectively. Feedback could be from educators, peers or
GenAI. The approach to feedback opportunities, particularly those using GenAI,
should be discussed at the start of the process and revisited/evaluated to
determine their effectiveness. Discuss how GenAI might be used to provide
feedback and the importance of reflecting on suggested improvements.

54


8. Discuss appropriate and transparent use of GenAI making clear that support
should complement, not replace students’ thinking and academic judgement.
Explicitly teach students how AI tools can be used ethically and within academic
regulations. Consider using the Generative AI CHECKlist cards (Beckingham &
Hartley, 2025) to engage students in an active discussion about the appropriate
use of AI.
Case study summary: Assessment – A Multimodal Approach to Unpacking a
Dissertation Assessment Brief by Sue Beckingham, Associate Professor, Sheffield Hallam
University

This case study outlines how an AI generated infographic maker called Napkin.AI was
used to create a collection of visuals based on information provided to it by students. The
prompts given to Napkin.AI were to create the infographics and NotebookLM was used to
create a podcast discussion based on the infographics.
(Full case study is included in Appendix A.)

55


Case study summary: Assessment - Formative Assessment Activities for Using
Generative AI Responsibly and Ethically by Sue Beckingham, Associate Professor,
Sheffield Hallam University
The Generative AI CHECKLIST is a multimodal resource that can be used as a formative
assessment activity to discuss how this technology can be used responsibly and ethically.
It uses an infographic poster to outline nine points for consideration which can be used as
a handout. The infographic PDF was then used to create a NotebookLM podcast which
discusses what the checklist is and the benefits of using it. This audio recording can then
be shared with students as a summary of the checklist. To check understanding, ChatGPT
is used to create a multiple-choice quiz.
(Full case study is included in Appendix A.)

Case study summary: Assessment – Strategies for Integrating AI-generated (GenAI)
Technologies in Assessment by Mari Cruz García Vallejo, Digital Education Consultant
and Affiliated Lecturer at the ULPGC (Spain)
This bilingual case study presents a practical design guide intended for both novice and
experienced higher education lecturers. It provides strategies for integrating AI-generated
(GenAI) technologies in assessment within their teaching practices. The design guide is
based on the concept of authentic assessment, a practical approach that enables
students to connect and apply knowledge gained in the classroom to real-world tasks or
projects relevant to their academic or professional disciplines. Additionally, the guide
promotes assessment for learning, particularly through the modality of assessment as
learning, by shifting the assessment focus to students themselves, who take on the roles
of assessors of their own work and that of their peers.
(Full case study is included in Appendix A.)

6.3.1.2 AI for Building/Transforming Assessments
When building/transforming multimodal assessments using GenAI, you should take the
following steps:
1.

Evaluate the learning outcome to which the assessment is/will be tied. What skill,
knowledge, or ability are you seeking to assess, and what do you think is the most
appropriate active task that students could do to show you that they have met
that skill?

56


2.

Consider the different stages, elements or aspects of the task you have set. The
focus of the overall task will determine which of these stages is the most
pronounced or vital to the assessment as a whole, giving you something to target.
For example, is the assessment primarily focused on the research process? Or is it
the students’ ability to present information? Or is it to test their capacity to analyse
an artefact?

3.

In what way is the assessment multimodal, and/or could it incorporate
multimodality to help them achieve the goal more effectively and/or get more out
of the assessment task?

4. Ensure that multimodality is used in conjunction with the correct learning
environment and with the correct resources so that students are able to complete
the task effectively, and in a format that avoids creating either cognitive or
assessment overload (Varga-Atkins, 2024).
5. Connect the two concepts together – the focus of the task and its multimodal
format. How might the focus of the task be presented multimodally? For example, if
the task focuses on the students’ ability to research a topic, can they produce a
pictorial representation of the landscape of the field as a result of their work?
6. The focus of the task and the multimodal elements within it provide steerage on
the way GenAI might be helpful for students. The use of GenAI can take three forms.
6.1. You can consider whether there are GenAI tools that might help with creating
certain elements of the overall submission that would normally take too large
an amount of time and effort to create relative to their importance to the
assessment as a whole (such as images, videos, audio recordings,
infographics, etc.)...
6.2. …Or you might consider whether GenAI might be useful in either the research
process or the dissemination/presentation activities, if applicable. AI might
help students to construct literature reviews, bibliographies or pictorial
representations of the field, create multimodal artefacts that present the
findings or overall message of their artefact.
6.3. Thirdly, you might consider whether GenAI might act as an additional source
of support or information for students completing the assessment. Can GenAI
be used to create datasets on which students perform analyses? Or can it act
as a character or figure with which the student interacts to help complete the
assessment?
7.

Regardless of how GenAI is positioned in the multimodal assessment, you should
provide the instructions on the assessment, and the associated use of GenAI, in
clear terms to students via assessment briefs, marking criteria, rubrics, etc. These

57


can also be multimodal for maximum accessibility, and GenAI might help
reformulate this information into more accessible or user-friendly formats.
8. Instructions for the use of GenAI in the assessment context should always include
reference to specific tools that meet institutional requirements and/or adhere to
institutional policies on the use of the technology.
9. Instructions for the use of GenAI should also specifically reference ways to use the
technology relevantly and ethically, and should instruct students to limit their use
of the technology to contexts where it is actually needed on sustainability grounds.
10. Any multimodal assessment designed that incorporates GenAI in some way
should also be able to be completed without the use of GenAI, to allow students
who do not wish to use it on ethical grounds, or otherwise are unable to use it, to
complete the task just as effectively as those who did. Those who use GenAI should
not be given an advantage over those who choose not to.
Case study summary: Assessment - How much can AI (at present) Contribute to
Identifying and Writing about Classification and Nature? by Tom Hartman, Lecturer in
Biology, University of Nottingham – Biology
This case study outlines an AI-enabled approach to accurately identifying wildlife species
via photographs and to develop critical appraisal of Generative AI outputs as part of the
process. To supplement students’ training in taking high-quality wildlife photographs and
accurately identifying species, they were given briefs to identify birdsong using the Merlin
app, take images of different species and identify them using Google Lens and then
critically evaluate text on the locality and wildlife written by ChatGPT. They were then
asked to mark the AI derived essay and write a reflection on the experience.
(Full case study is included in Appendix A.)

6.3.1.3 AI for Completing Assessments
When students are completing multimodal assessments using GenAI, they should
consider:
1.

The most appropriate point at which GenAI can be used so that it is at its most
helpful. For example, GenAI could be very helpful for creating background assets or
imagery that would take a lot of time relative to their importance to the overall
assessment.

2.

The position of GenAI at every stage of the assessment process, relative to the
instructions of the assessment task:
a. In the initial stages, GenAI could be used as a tutor or research assistant to
refine research questions, scaffold the approach to the assessment in the
58


context of the marking criteria, or perform and evaluate initial
research/understand the critical field and how it interacts.
b. In the completion stage, GenAI could be used as an assistive technology to
help construct different aspects of the multimodal assessment. GenAI can
be very helpful to create individual aspects of a multimodal artefact, but it
is up to the student to ensure that the multimodal components are
integrated together to form the coherent whole in a way that makes sense,
that answers the task of the assessment, and says what the student wants
it to say.
c. In the post-assessment phase (or perhaps the post-completion, presubmission phase), GenAI can be used again as a reflective tutor or
assistant to reflect on how well the multimodal submission answers the task
of the assessment and to compare the submission to the structural aspects
of the assessment - marking criteria, rubrics, etc., in order to obtain
informal, additional feedback either before or after submission.
3.

How GenAI might be useful in turning a single-mode artefact into a multimodal
one – if a student has worked entirely textually to create a draft, GenAI can
perhaps help convert it into multimodal contexts, or vice versa.

4. How GenAI might be helpful in some of the wider aspects of the assessment that
do not directly relate to its purpose or argument, but which could still be useful –
for example, (re)formatting reference lists (which should be manually checked).

6.3.2 Feedback - Enhancing feedback practices
When considering using GenAI to enhance feedback practices, you should take the
following steps:
1.

Consider the context that you’re thinking about using GenAI in the specific
assessment that you have set students. Do you want AI to construct the feedback,
or reformulate it into alternative forms? Or both?

2.

Consider how you might encourage students to self-feedback using GenAI, or to
speed up self-feedback processes that already exist, like revision. For example,
students might consider making flashcards using lecture material and/or their
assessments and use them for exam revision, or they might submit formative
assessment tasks to a GenAI and ask it for feedback in multimodal forms.

3.

Consider how you might help develop students’ assessment literacy by allowing
them to mark a GenAI-generated version of the assessment before or after
completing it for themselves. This can be helpful as it would not be real student
work, can be calibrated according to the level at which you want the sample
assessment to be marked (i.e. asking GenAI to construct a ‘good’ piece of work or a

59


‘poor’ one), and can be structured to also show students specifically what their
assessment could look like when they complete it for themselves.
4. Consider using AI-enabled feedback processes throughout the assessment - an AI
reflection log, for example, can help students make effective use of the technology,
reflect on their assessments, and also critically consider GenAI itself as they
complete the assessment.
5. Consider the use of GenAI in your own processes – it can perhaps help you to
reframe some of your language into more accessible formats, pictorial formats, or
other multimodal contexts (and you can discuss your own reactions with
students).
Case study summary: Assessment – Using NotebookLM to Develop Students’
Assessment and Feedback Literacy by Samuel Saunders, Educational Developer,
University of Liverpool
This case study explores how students can use NotebookLM, a Google product that is
designed as a research organiser, note-taker and resource manager, to develop their
assessment and feedback literacy. Students use the ‘audio overview’ feature of the
platform to generate an AI-presented podcast of their draft assessments, on which they
then reflect and use to enhance the draft ahead of summative submission. Students
upload their work to NotebookLM, generate the podcast and listen to the output, and can
interact with the AI presenters to determine what the assessment is doing well in terms of
argument, structure, analysis or evaluation, etc., Students can ask the presenters what
they need to do next to ensure it reaches the higher levels of the marking criteria/rubric.
The activity is designed to both develop their literacy around their own work and what
they need to do to succeed in the assessment itself. It also provide a new space for
dialogic interaction concerning an assessment and feedback process.
(Full case study is included in Appendix A.)

60


7. Considerations
In this section, we explore the challenges and risks of using AI (7.1) as well as selecting
GenAI platforms for multimodal contexts (7.2).

7.1 Challenges and risks of using GenAI in Learning, Teaching
and Assessment
Above, we have listed some of the benefits of using GenAI in multimodal education. In this
section, we enlist some of its challenges, complementing our earlier Section 2 on
Responsible use of AI.

7.1.1 Risks of stifling creativity
While GenAI can enhance creativity by generating diverse ideas and formats, there is also
a risk that over-reliance on GenAI outputs may constrain original thinking. If students
default to AI-generated suggestions without critically interrogating or reworking them,
they may miss opportunities to develop their own creative problem-solving skills (Chan &
Hu, 2023). Educators should work to explicitly design activities that require students to
extend, adapt, critique, or even abandon, GenAI outputs rather than adopt them
wholesale.

7.1.2 Potential for dependency on GenAI
The convenience of AI tools may inadvertently foster dependence, where students come
to expect AI to initiate or structure their work. This can undermine the development of
essential skills such as independent research, critical analysis, and self-regulation
(Beckingham et al., 2024). Clear guidance is needed on when and how GenAI should
support learning tasks, making space for students to engage with content and ideas
without technological mediation.

7.1.3 Bypassing reflective processes
GenAI’s speed and efficiency can compress stages of the learning cycle that are vital for
reflection and metacognition (Ng et al., 2024). For instance, automated summaries or
visualisations may provide answers too quickly, bypassing the deep thinking that arises
from manual synthesis. Educators should intentionally build pause points into tasks that
prompt students to reflect on both the content and the process of working with AI,
ensuring that reflective learning is not lost amidst efficiency gains.

7.2 Selecting GenAI platforms
Selecting and using generative AI (GenAI) platforms for multimodal teaching, learning
and assessment requires balancing pedagogical purpose, ethical practice and practical

61


functionality. This section provides principles and starting points rather than specific
recommendations, recognising that platforms evolve rapidly.

7.2.1 Purpose-driven selection
Before choosing a tool, consider:
•

What is the learning outcome? Clarify whether the goal is idea generation,
multimodal content creation (e.g. text-to-image, text-to-audio), or critical
evaluation of AI outputs.

•

Which modalities are essential? Decide whether visuals, audio, video or interactive
simulations are genuinely needed to meet the outcome.

•

Who are the users? Reflect on students’ digital skills, accessibility needs and device
access.

•

Is it ethical and inclusive? Check for potential bias, privacy and representation
issues in the platform’s design and outputs.

7.2.2 Typology of multimodal GenAI tools
Drawing on the There is an App for that typology (see for instance the ‘There’s an AI for
that’ website), multimodal GenAI tools can be grouped by their core function. These
categories help educators map tools to learning outcomes without prescribing specific
platforms:
•

Text generation and enhancement – for producing written outputs, summaries or
scripts (e.g. large language models).

•

Text-to-image generation – for creating diagrams, visual metaphors or illustrative
content.

•

Text-to-audio or speech synthesis – for podcasts, narration or multimodal
feedback.

•

Image-to-text or multimodal analysis – interpreting images, diagrams or
handwriting.

•

Video and animation creation – for explainer videos, avatars or short animations.

•

Music and sound generation – for background tracks, soundscapes or audio cues.

•

Translation and language adaptation – converting or adapting content across
languages and modes.

•

Drawing and sketch-based tools – generating visuals from hand-drawn inputs
(e.g. concept sketches or quick diagrams).

•

Avatar and persona creation – developing characters or role-play scenarios to
enhance interactivity.

62


7.2.3 Steps to get started
1.

Check institutional policy: confirm approved tools and clarify data privacy
requirements (e.g. avoid uploading identifiable student work to public platforms).

2.

Create a secure account: use university logins where available and review privacy
settings to manage data retention.

3.

Start with one modality: begin by experimenting with one function (e.g. text-toimage) before layering multiple modes.

4. Model and scaffold usage: demonstrate your workflow to students, including
prompt design and critical evaluation of AI outputs.
5. Embed reflection and critique: incorporate activities where students assess outputs
for bias, accuracy and appropriateness for their discipline.

7.2.4 Key considerations
•

Bias and representation: While many platforms are improving (e.g. more diverse
image outputs), educators must still discuss and critique representation with
students.

•

Accessibility and inclusion: Multimodal formats can support neurodivergent and
multilingual learners but should remain optional.

•

Sustainability: High-energy tasks (e.g. image and video generation) should only be
used where they add pedagogical value.

•

Equity of access: Ensure tasks do not disadvantage students without access to
premium features or high-spec devices.

This guide has avoided endorsing particular tools in its main sections to remain relevant
as the technology evolves. Educators are encouraged to regularly review and share
emerging tools within institutional communities of practice.

63


8. Towards a Unifying Model of Multimodal Learning
Design with Generative AI (MMLD-AI)
This guide has somewhat compartmentalised Generative AI in multimodal pedagogy by
organising it in discrete sections on learning, teaching and assessment. We have done so
for the sake of focus rather than to position these as separate entities - they are always
interconnected and interdependent. Overall, we have striven to point out how Generative
AI might be useful, helpful, or fundamentally embedded into pedagogic activities across
all three spaces, and offer advice, guidance and tips for implementing it alongside case
studies where others have already done so. However, there is natural overlap between
these three areas, and so it makes sense to consider how Generative AI might be
embedded into all areas of pedagogy simultaneously and whether there may be a
unifying model or way to express this.
To help facilitate this unifying approach, we contend that educators can consider
multimodal learning design with GenAI guided by both the principles of Universal Design
for Learning (CAST 2024), and the ABC Design for Learning approach.
Universal Design for Learning (UDL) argues that learning environments should be flexible,
should accommodate the diverse needs of all students, and should focus on developing
students’ agency over their own learning and/or that learning should be self-constructed.
Its three principles are designing options for 1) multiple means of engagement, 2) multiple
means of representation and 3) multiple means of action and expression.
The ABC Design for Learning approach, meanwhile, storyboards the student’s journey
through a module/programme and helps educators consider the types of learning
activities that students would engage with in order to progress in their learning, relating to
learning purpose, structure and outcome. For more information, please consult the ABC
Learning Design (Laurillard 2012; Young & Peroviċ, 2019).
The figure below (Figure 6) shows how these two designs have been adapted and
merged into a unifying model of multimodal learning design with Generative AI (MMLDAI). On the left, the model indicates that education actors need to consider multimodality
when planning and engaging in: 1) instruction, 2) learning resources/representations, and
3) student action and expression, taking into account the principles of Universal Design for
Learning. The model then zooms in on students’ action and expression, focusing on six
multimodal engagement types with Generative AI on the right, adapted from the six
learning types or acts of the ABC Learning Design. In other words, it can provide a
‘storyboard’ of the six types of educational engagements with GenAI, in relation to the
three interconnected areas of education shown on the left, with emphasis on student
multimodal action and expression:

64


•

acquisition of information,

•

investigation and/or research,

•

collaboration with others,

•

production of artefacts (for learning, teaching or assessment purposes),

•

practice of approaches/theories/principles/skills, and

•

discussion/discourse, including critique/evaluation.

Figure 6. MMLD-AI model: A unifying model of multimodal learning design with GenAI

Then, once the nature of learning type has been decided, [decision making – then decide
on strengths] multimodal affordances of GenAI and respective strengths of the cybersocial learning are to be explored. The MDL-AI model also includes different processes,
building on the work by Galla, Cope & Kalantzis (2025), which revisits Bloom’s taxonomy.
We recommend that their Dual-Track Cyber-Social Learning Model is considered as
complementary to the MDL-AI model, as it clearly highlights the strengths of both humans
and AI in learning designs.
Galla, Cope and Kalantzis (2025) detail (Table 1) the learning process in relation to human
vs AI strengths (e.g. in interpretation and analysis) and based on these how an effective
cyber-social approach may look like that utilise the strengths of each (human/AI),which
they call the dual-track cyber social model. The authors also suggest that effective
learning design should cover key areas of exploring effective AI/human collaboration
strategies (prompting, evaluation, verification), fostering metacognitive awareness (self
vs AI strengths/weaknesses), and AI literacy, i.e. responsible, ethical use of GenAI (bias,
privacy).

65


Table 1. The Dual-Track Cyber-Social Learning Model (Galla et al., 2025, pp 15-16): Human–
AI strengths and cyber-social approach across processes
Process

Human strengths

AI strengths

Cyber-social approach

Knowledge &

Contextual

Rapid data retrieval

Humans define creative

framing

understanding;

across vast

purpose and frame

Purpose-driven

sources; Statistical

problems; AI generates

inquiry; Common

pattern recognition;

background data and

sense & embodied

Broad topical

inspiration; Humans

knowledge; Critical

coverage

filter for relevance and

verification

verify information

Interpretation

Causal reasoning;

Correlation

AI identifies statistical

& analysis

Cultural and ethical

analysis;

patterns and

awareness; Implicit

Theme/pattern

correlations; Humans

meaning recognition;

identification;

determine causality,

Emotional intelligence

Feature extraction

relevance, and deeper

at scale;

significance; Together

Systematic

they build richer

decomposition

understanding

Application &

Situated judgement;

Rapid simulation of

AI rapidly generates

prototyping

Adaptive problem-

multiple scenarios;

digital prototypes or

solving; Ethical

Consistent rule

models; Humans adapt

decision-making;

application;

designs for real-world

Craftsmanship

Code/digital

complexity, apply

artefact generation

physical craft, and
ensure ethical
implementation

Synthesis &

Novel conceptual

Cross-domain

AI explores possible

creation

blending; Purpose-

pattern integration;

combinations and

driven integration;

Combinatorial

variations; Humans

Narrative coherence;

exploration; Rapid

provide creative vision,

Embodied insights

iteration of

ensure conceptual

variations

coherence, infuse
emotional depth, and
establish authentic
voice

66


Evaluation &

Ethical and aesthetic

Quantitative

AI provides objective

refinement

judgement;

assessment;

metrics and

Contextual

Consistency

consistency checks;

appropriateness;

checking;

Humans make value

Holistic value

Comparison across

judgements, assess

assessment; Impact

cases; Anomaly

ethical implications,

consideration

detection

and determine
significance

Together with the MDL-AI model, this table provides a means of tying together all the
approaches outlined in the Teaching, Learning and Assessment/Feedback sections of this
guide. If it is implemented sensibly, critically, and as ethically as possible, GenAI in
multimodal contexts can help students achieve all six of these steps in effective ways,
across all three pedagogic contexts.
One way of progressing Galla et al.’s cyber-social dual-track model would be to refine
and further develop it for multimodal GenAI contexts. With emerging technologies in areas
such as touch perception, dexterity, human–robot interaction, as well as GenAI and
multimodal data, embodied intelligence is expected to evolve further, supporting richer
interactions between students and computers in both virtual and physical environments
(Feng et al., 2025).

67


9. Reference List
Alasadi, E. A., & Baiz, C. R. (2024). Multimodal generative artificial intelligence tackles visual
problems in chemistry. Journal of Chemical Education, 101(7), 2716–2729.
https://doi.org/10.1021/acs.jchemed.4c00138
Beckingham, S., & Hartley, P. (2025a). In search of ‘Responsible’ Generative AI (GenAI). In M.
A. Doolan & L. Ritchie (Eds.), Transforming teaching excellence: Future proofing
education for all. Leading Global Excellence in Pedagogy, Volume 3. IFNTF
Publishing.
Beckingham, S., & Hartley, P. (2025b). The Generative AI CHECKLIST discussion cards
[Presentation]. National Teaching Repository.
https://doi.org/10.25416/NTR.29410013.v1
Beckingham, S., Lawrence, J., Powell, S., & Hartley, P. (2024). Using generative AI effectively
in higher education: Sustainable and ethical practices for learning, teaching and
assessment. Routledge.
Bolick, A. D., & da Silva, R. L. (2024). Exploring artificial intelligence tools and their potential
impact to instructional design workflows and organizational systems. TechTrends,
68(1), 91–100. https://doi.org/10.1007/s11528-023-00894-2
Bond, M., Khosravi, H., De Laat, M., Bergdhal, N., Negrea, V., Oxley, E., Pham, P., Wang Chong,
S., & Siemens, G. (2024). A meta systematic review of artificial intelligence in higher
education: A call for increased ethics, collaboration, and rigour. International
Journal of Educational Technology in Higher Education, 21(4).
https://doi.org/10.1186/s41239-023-00436-z
CAST. (2024). Universal Design for Learning Guidelines version 3.0. Retrieved August 19,
2025, from https://udlguidelines.cast.org
Chan, C. K. Y., & Hu, W. (2023). Students’ voices on generative AI: Perceptions, benefits, and
challenges in higher education. International Journal of Educational Technology in
Higher Education, 20(43). https://doi.org/10.1186/s41239-023-00411-8
Chen, Y., Kumar, A., & Li, X. (2023). Exploring possibilities of AI-enabled image synthesis and
design in education. Proceedings – Frontiers in Education Conference (FIE).
https://doi.org/10.1109/FIE58773.2023.10342958
Cope, B., & Kalantzis, M. (2024). A multimodal grammar of artificial intelligence: Measuring
the gains and losses in generative artificial intelligence. Multimodality & Society,
4(2), 123–152. https://doi.org/10.1177/26349795231221699
68


Feng, S., Zhang, H., & Gašević, D. (2025). Where is AIED headed? Key topics and emerging
frontiers (2020-2024) (arXiv:2506.20971). arXiv.
https://doi.org/10.48550/arXiv.2506.20971
Fui-Hoon Nah, F., Zheng, R., Cai, J., Siau, K., & Chen, L. (2023). Generative AI and ChatGPT:
Applications, challenges, and AI-human collaboration. Journal of Information
Technology Case and Application Research, 25(3), 277–304.
https://doi.org/10.1080/15228053.2023.2233814
Furze, L. (2024). Practical AI strategies. Amba Press.
Galla, M., Cope, B., & Kalantzis, M. (2025). Generative AI turns Bloom’s Taxonomy upsidedown: Implications for creativity and learning. In Oxford handbook of human
creativity x generative AI in education. Oxford University Press.
Gibson, J. J. (2014). The theory of affordances. In The people, place, and space reader (pp.
56–60). Routledge. (Original work published 1979)
Hall, R. (2024). Generative AI and re-weaving a pedagogical horizon of social possibility.
International Journal of Educational Technology in Higher Education, 21(1).
https://doi.org/10.1186/s41239-024-00445-6
Heilala, V., Araya, R., & Hämäläinen, R. (2025). Beyond text-to-text: An overview of
multimodal and generative artificial intelligence for education using topic
modelling. ArXiv. https://arxiv.org/abs/2409.16376v2
Henderson, M., Selwyn, N., & Aston, R. (2017). What works and why? Student perceptions of
‘useful’ digital technology in university teaching and learning. Studies in Higher
Education, 42(8), 1567–1579. https://doi.org/10.1080/03075079.2015.1007946
Hsiao, M.-Y., & Zhang, S. (2023). Research on the educational application of generative
artificial intelligence images in the design of semiotics learning models. In
Proceedings of the 2023 6th International Conference on Educational Technology
Management (pp. 8–15). Association for Computing Machinery.
https://doi.org/10.1145/3637907.3637947
Hung, H.-T., Chiu, Y.-C. J., & Yeh, H.-C. (2012). Multimodal assessment of and for learning: A
theory-driven design rubric. British Journal of Educational Technology, 44(3), 400–
409. https://doi.org/10.1111/j.1467-8535.2012.01337.x
Kalantzis, M., & Cope, B. (2025). Literacy in the time of artificial intelligence. Reading
Research Quarterly, 60(1), https://doi.org/10.1002/rrq.591

69


Kim, J., Detrick, R., Yu, S., Song, Y., Bol, L., & Li, N. (2025). Socially shared regulation of learning
and artificial intelligence: Opportunities to support socially shared regulation.
Education and Information Technologies. https://doi.org/10.1007/s10639-024-131879
Kim, J., Lee, S.-S., Detrick, R., Wang, J., & Li, N. (2025). Students-generative AI interaction
patterns and its impact on academic writing. Journal of Computing in Higher
Education. Advance online publication. https://doi.org/10.1007/s12528-025-094446
Kofinas, A., Tsay, C., & Pike, D. (2025). The impact of generative AI on academic integrity of
authentic assessments within a higher education context. British Journal of
Educational Technology, 1–28. Advance online publication.
https://doi.org/10.1111/bjet.13585
Koppam, M., Mehta, R., Kulkarni, R., Pai, S., & Srinivas, K. S. (2024). Edvent—Speech to
presentation automation. In A. Iglesias, J. Shin, B. Patel, & A. Joshi (Eds.), Lecture
Notes in Networks and Systems, 834, (pp. 305–325). Springer.
https://doi.org/10.1007/978-981-99-8349-0_24
Kress, G. (2009). Multimodality: A social semiotic approach to contemporary
communication. Routledge.
Kress, G., & Van Leeuwen, T. (2001). Multimodal discourse: The modes and media of
contemporary communication. Oxford University Press.
Kress, G., & Van Leeuwen, T. (2020). Reading images: The grammar of visual design.
Routledge.
Lacković, N. (2020). Inquiry graphics in higher education: New approaches to knowledge,
learning and methods with images. Springer. https://doi.org/10.1007/978-3-03039387-8
Lacković, N., & Olteanu, A. (2023). Relational and multimodal higher education: Digital,
social, and environmental perspectives. Taylor & Francis.
https://doi.org/10.4324/9781003155201
Laurillard, D. (2012). Teaching as a design science: Building pedagogical patterns for
learning and technology. Routledge.
Leoste, J., Fenyvesi, K., & di Rosario, G. (2025). Socratic dialogue with generative artificial
intelligence: Where is the future? In K. Arai (Ed.), Advances in information and

70


communication. FICC 2025 (Lecture Notes in Networks and Systems, Vol. 1283).
Springer. https://doi.org/10.1007/978-3-031-84457-7_20
Li, N., Huijser, H., Zhang, S., Zhang, T., & Zhang, X. (2024). Putting the flesh on the bones: Using
a fishbone digital learning design method to align educational objectives.
Education and Information Technologies. https://doi.org/10.1007/s10639-02412799-5
Lim, F. V. (2020). Designing learning from embodied teaching: Perspectives from
multimodality. Routledge.
Lim, F. V., & Tan-Chia, L. (2023). Designing learning for multimodal literacy: Teaching
viewing and representing. Routledge.
Luccioni, A. S., Jernite, Y., & Strubell, E. (2024a). Power hungry processing: Watts driving the
cost of AI deployment? In Proceedings of the 2024 ACM Conference on Fairness,
Accountability, and Transparency (pp. 85–99). Association for Computing
Machinery. https://doi.org/10.48550/arXiv.2311.16863
Martin, D. (2020). Providing students with multimodal feedback experiences. Journal of
Curriculum, Teaching, Learning, and Leadership in Education, 5(1).
https://doi.org/10.32873/uno.dc.ctlle.05.01.1082
Middleton, A., & Nortcliffe, A. (2010). Audio feedback design: Principles and emerging
practice. International Journal of Continuing Engineering Education and Life-Long
Learning, 20(2), 208–223. https://doi.org/10.1504/ijceell.2010.036816
Morris, C. (2025, May 28). Anthropic CEO warns AI could eliminate half of all entry-level
white-collar jobs. Fortune. https://fortune.com/2025/05/28/anthropic-ceowarning-ai-job-loss/
Naughton, J. (2025, June 27). Does AI make us dull? Students have answered that question.
Observer. https://observer.co.uk/news/science-technology/article/does-aimakes-us-dull-students-have-answered-that-question
Ng, D. T. K., Xinyu, C., Leung, J. K. L., & Chu, S. K. W. (2024). Fostering students’ AI literacy
development through educational games: AI knowledge, affective and cognitive
engagement. Journal of Computer Assisted Learning, 40(5), 2049–2064.
https://doi.org/10.1111/jcal.13009
Nortcliffe, A., & Middleton, A. (2012). Smartphone feedback: Using an iPhone to improve the
distribution of audio feedback. International Journal of Electrical Engineering &
Education, 48(3), 280–293. https://doi.org/10.7227/IJEEE.48.3.6
71


Pinski, M., & Benlian, A. (2024). AI literacy for users – A comprehensive review and future
research directions of learning methods, components, and effects. Computers in
Human Behavior: Artificial Humans, 2(1), 100062.
https://doi.org/10.1016/j.chbah.2024.100062
Ross, J., Curwood, J. S., & Bell, A. (2020). A multimodal assessment framework for higher
education. E-Learning and Digital Media, 17(4), 290–306.
https://doi.org/10.1177/2042753020927201
Rusconi, L., & Squillaci, M. (2023). Effects of a universal design for learning (UDL) training
course on the development teachers’ competences: A systematic review.
Education Sciences, 13(5), 466. https://doi.org/10.3390/educsci13050466
Saunders, S., Coulby, C., & Lindsay, R. (2024). Pedagogy and policy in a brave new world: A
case study on the development of generative AI literacy at the University of
Liverpool. In S. Beckingham & P. Hartley (Eds.), Using generative AI effectively in
higher education. Routledge. https://doi.org/10.4324/9781003482918-3
Tan, X., Xu, W., & Wang, C. (2025). Purposeful remixing with generative AI: Constructing
designer voice in multimodal composing. Computers and Composition, 75, 102893.
https://doi.org/10.1016/j.compcom.2024.102893
Tay, J., Li, N., Luo, L., Zhou, Z., Meng, W., Zhang, Q., Purwanto, E., & Lu, Y. (2025). Role of learner
agency for interactive group learning through the lens of Blooms Taxonomy.
Interactive Learning Environments. https://doi.org/10.1080/10494820.2025.2507277
Tomlinson, K., Jaffe, S., Wang, W., Counts, S., & Suri, S. (2025). Working with AI: Measuring the
occupational implications of generative AI. ArXiv. https://arxiv.org/abs/2507.07935
Tzirides, A. O., Zapata, G., Kastania, N. P., Saini, A. K., Castro, V., Ismael, S. A., You, Y., Santos, T.
A. D., Searsmith, D., O’Brien, C., Cope, B., & Kalantzis, M. (2024). Combining human
and artificial intelligence for enhanced AI literacy in higher education. Computers
and Education Open, 6. https://doi.org/10.1016/j.caeo.2024.100184
Vallor, S. (2024). The AI mirror: How to reclaim our humanity in an age of machine thinking.
Oxford University Press. https://doi.org/10.1093/oso/9780197759066.001.0001
Van Leeuwen, T. (2017). Multimodal literacy. Viden Om Laesning, 21, 4–11,
https://videnomlaesning.dk/media/2127/21_theo-van-leeuwen.pdf.
Varga-Atkins, T. (2024). Multimodal learning: A practitioner guide. AdvanceHE.
https://www.advance-he.ac.uk/knowledge-hub/multimodal-learningpractitioner-guide
72


Varga-Atkins, T., & O’Brien, M. (2009). From drawings to diagrams: Maintaining researcher
control during graphic elicitation in qualitative interviews. International Journal of
Research & Method in Education, 32(1), 53–67.
https://doi.org/10.1080/17437270902759998
Voelkel, S., & Mello, L. (2015). Audio feedback – better feedback? Bioscience Education,
22(1), 16–30. https://doi.org/10.11120/beej.2014.00022
Walter, Y. (2024). Embracing the future of artificial intelligence in the classroom: The
relevance of AI literacy, prompt engineering, and critical thinking in modern
education. International Journal of Educational Technology in Higher Education,
21(1). https://doi.org/10.1186/s41239-024-00448-3
Welsh, K., & Milne, L. (2025, June). Artificial intelligence and sustainability [Presentation].
Generative AI Network Webinar, Online.
Yan, L., Echeverria, V., Jin, Y., Fernandez-Nieto, G., Zhao, L., Li, X., Alfredo, R., Swiecki, Z.,
Gašević, D., & Martinez-Maldonado, R. (2024). Evidence-based multimodal learning
analytics for feedback and reflection in collaborative learning. British Journal of
Educational Technology, 55(5), 1900–1925. https://doi.org/10.1111/bjet.13498
Yan, L., Greiff, S., Lodge, J. M., & Gašević, D. (2025). Distinguishing performance gains from
learning when using generative AI. Nature Reviews Psychology, 4, 435–436.
https://doi.org/10.1038/s44159-025-00467-5
Young, C., & Perović, N. (2016). Rapid and creative course design: As easy as ABC? In J.
Domenech, M. Vincent-Vela, R. Peña-Ortiz, E. de la Poza, & D. Blazquez (Eds.),
Procedia - Social and Behavioral Sciences, 228, (pp. 465-470). Elsevier.
Yun, B., Feng, D., Chen, A., Nikzad, A., & Salehi, N. (2025). Generative AI in knowledge work:
Design implications for data navigation and decision-making. In Proceedings of
the CHI Conference on Human Factors in Computing Systems. Association for
Computing Machinery. https://doi.org/10.1145/3706598.3713337
Zawacki-Richter, O., Marín, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of
research on artificial intelligence applications in higher education – where are the
educators? International Journal of Educational Technology in Higher Education,
16(1), 39. https://doi.org/10.1186/s41239-019-0171-0

73


Appendix A. Case studies
Table of Case Studies
Case Study

Author(s)

Category

Using Generative AI to Enhance Learning

Sarvin Bahmani

Learning

Sue Beckingham

Assessment

Sue Beckingham

Assessment

Live Storytelling through Generative AI

Simon Campion, Radoslaw

Learning

Powered Avatars

Dorociak, Georg Meyer

Strategies for Integrating AI-generated

Mari Cruz García Vallejo

Assessment

Gareth Ellis

Learning

Katherine Geer, Daniel Hsu

Learning

Ioannis Glinavos

Teaching

Tom Hartman

Assessment

Ayaat Jaway

Learning

Liam Kaye, Chris Barlow

Learning

Rob Lindsay

Teaching

in Computational Chemistry
A Multimodal Approach to Unpacking a
Dissertation Assessment Brief
Formative Assessment Activities for Using
Generative AI Responsibly and Ethically

(GenAI) Technologies in Assessment
Experimental Practice – Using Mind Maps
in Project Planning to Foster Inclusion
Enhancing Learning with AI Tools for Idea
Generation
Enhancing Engagement through
Streaming Avatars and Synchronous AI
Assistants in Online Education
How much can AI (at present) Contribute
to Identifying and Writing about
Classification and Nature?
Using Generative AI to Aid Studying and
Learning for Medical Students
Exploratory Learning using GAI to support
Presentations
Peer Conflict Resolution in Group Work:
Using GenAI to Represent Subject
Knowledge Multimodally

74


A Multimodal Groupwork Activity:

Alice Maher

Learning

AI Visual Metaphor Activity (SEDA

Elora Marston, Jonathan

Teaching

Learning to Tutor Online Course)

Rhodes

Leveraging Generative AI for Enhanced

Magdalena Plesa

Learning

Samuel Saunders

Assessment

Perk Up Your Academic Journey: A Coffee

Laura Sharp, Eric Davies, Mia

Teaching

Shop Adventure

Wilson, Ailsa Foley

Improving Engagement with eLearning

Vivien Shaw

Teaching

Farkhondeh Vahdati

Learning

Fostering Belonging Among First-Year
Bioscience Students

Entrepreneurial Teaching
Using NotebookLM Generated Podcasts
to Develop Students’ Assessment and
Feedback Literacy

by Creating Multimodal Video-Explainer
The role of GenAI in Design for the
Disassembly of Emergency Shelters

75


Using Generative AI to Enhance Learning in Computational
Chemistry
Name: Sarvin Bahmani
Job Title and Affiliation: PhD Student, Computer Science, University of Liverpool
Contact Details: R.Bahmani@liverpool.ac.uk

What is the MML Idea?
I used multimodal Generative AI tools such as ChatGPT and DALL·E alongside
computational platforms like Comgen and FUSE. These tools allowed me to create
learning materials that combine textual summaries, chemical diagrams, interactive code,
and teaching presentations. By simplifying dense textbook theory and translating it into
accessible outputs, I aim to improve learning for diverse student audiences, especially
those who require more inclusive resources.
This example includes Generative AI elements of:
•

Image

•

Code

•

Text

•

Speech

Why This Idea?
Computational chemistry is inherently multimodal—it involves theoretical descriptions,
visual representations of chemical structures, and the practical use of code. Traditionally,
this creates barriers for students, particularly when material is abstract, inaccessible, or
overly technical.
To address this, I integrate multimodal GenAI in my work to break down chemical
concepts. I use NLP tools to extract and simplify content from textbooks—for example,
converting explanations of “ionic charge balancing” or “SMT solving” into clear, studentfriendly summaries. Visual learners benefit when I prompt tools to generate diagrams like
“a lithium-rich oxide lattice showing charge balancing,” making the invisible visible.
GenAI also helps me generate and debug code for chemistry software, simplifying
workflows and making advanced methods more approachable. In presentations, I use
GenAI to create concise, illustrated slides explaining topics such as constraint-based vs.
enumeration-based composition generation. Finally, I upload molecular diagrams to
receive automatically generated text annotations, improving accessibility for students
who struggle with visual processing.
By combining these outputs, I create inclusive and engaging learning experiences that
support student understanding across levels of familiarity.
76


How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
Step 1: Input complex chemistry text into GenAI for simplification
Use the natural language processing (NLP) capabilities of GenAI (e.g. GPT-4) to simplify
complex textbook explanations or technical terms.
Example prompt:
‘Simplify the explanation of ionic charge balancing in lithium-rich oxides for a
beginner student.’
Step 2: Generate a visual diagram from a text prompt
Use GenAI’s visual generation capabilities to turn abstract chemical concepts into
diagrams.
Example prompt:
‘Generate a diagram illustrating charge balancing within a lithium-rich oxide
lattice.’
Step 3: Input code snippets for explanation or debugging
If working with Python tools like FUSE or Comgen, paste code into the GenAI platform and
request clarification or fixes.
Example prompt:
‘Explain the following code used for SMT solving in a ternary system.’
Step 4: Request GenAI to create teaching slides
Use GenAI to develop slides that present complex information in a digestible and visually
engaging way.
Example prompt:
‘Create a clear slide explaining constraint-based versus enumeration-based
composition generation in chemistry.’
Step 5: Convert Images into Descriptive Text
Leverage image input features and upload chemistry-related images (e.g. molecular
diagrams) and use GenAI’s image recognition capabilities to produce text descriptions or
annotations.
Example prompt:
‘What is this diagram showing? Add labels and a description suitable for
undergraduate students.’

77


This approach can be adopted by educators or students in computational chemistry or
related fields to enhance learning through multimodal GenAI. Each step helps make
abstract chemical concepts more accessible, supporting diverse learners and
encouraging engagement through visual and interactive formats.

Links to Tools and Resources:
•

ChatGPT

•

Gemini

Image:

Figure 1. Automatically Generated Textual Annotation from Chemical Structure Image (output
example)

Using GenAI’s multimodal capabilities, an uploaded chemical reaction diagram is
analysed and clearly annotated:
‘This diagram shows the decomposition reaction of ammonium dichromate
(NH4)2Cr2O7(NH_4)_2Cr2O__7(NH4)2 Cr2 O7. It breaks down into chromium(III) oxide
Cr2O3Cr2O__3Cr2 O3, nitrogen gas N2N_2N2, and water vapor H2OH2OH2 O. The visual
clearly represents each molecule's structure: ammonium ions NH4+NH
4^+NH4+, dichromate_ions Cr2O72−Cr2O__7^{2-}Cr2 O72−, and the products formed.
Such annotated explanations help students directly interpret chemical visuals, making
complex chemical reactions easier to understand.’
This type of annotated visual greatly improves accessibility for students who benefit from
clear, multimodal explanations of chemical structures and reactions.

References:
Clymo, J., Collins, C.M., Atkinson, K., Dyer, M.S., Gaultois, M.W., Gusev, V.V., ... & Schewe, S.
(2025) Exploration of chemical space through automated reasoning. Angewandte
Chemie, e202417657. Available at: https://doi.org/10.1002/ange.202417657

78


Collins, C., Darling, G.R. and Rosseinsky, M.J. (2018) The Flexible Unit Structure Engine (FUSE)
for probe structure-based composition prediction. Faraday Discussions, 211, pp.117–
131. Available at: https://doi.org/10.1039/C8FD00045J
Gusev, V.V., Adamson, D., Deligkas, A., et al. (2023) Optimality guarantees for crystal
structure prediction. Nature, 619, pp.68–72. Available at:
https://doi.org/10.1038/s41586-023-06071-y

79


A Multimodal Approach to Unpacking a Dissertation Assessment
Brief
Name: Sue Beckingham
Job Title and Affiliation: Associate Professor, Sheffield Hallam University
Contact Details: Sheffield Hallam University
Relevant Links:
Sheffield Hallam University - Sue Beckingham

LinkedIn - Sue Beckingham
What is the MML Idea?
I used an AI generated infographic maker called Napkin, which created a collection of
visuals based on the information provided. The prompts given to Napkin were to create
the infographics and NotebookLM was used to create a podcast discussion based on the
infographics.
This example includes Generative AI elements of:
•

Image

•

Sound

•

Speech

•

Text

Why This Idea?
The dissertation is traditionally a final capstone project and possibly the biggest written
assessment students will experience. Typically the students will receive an assessment
brief and assessment criteria which will be used as a marking rubric. Whilst this can be
presented as a very detailed Word document, it can be helpful to break up what is
required into smaller chunks and summarise with visuals. Infographics can convey
messages in a concise and engaging way, bringing out key points visually (Ritchie et al,
2012) Data visualisation can grab a reader’s attention through four visual preattentive
attributes: form, colour, spatial position and movement (Ware, 2020, 2021). For example, a
simple pie chart can be used to visualise data and colour to highlight different aspects.

How Could Others Implement This Idea?
The prompts given to Napkin were to create infographics based on the following headings
and bullet points:

80


Pre-Writing

Post-Writing

•

Topic Selection

•

Proofreading

•

Literature Review

•

Grammar check

•

Research existing studies

•

Formatting compliance

•

Identify research gap

•

Plagiarism Check

•

Research Design

•

Methodology

•

Ethical Approvals

Writing

Submission
•

Adhere to deadline

•

Upload as per guidelines

Evaluation

•

Introduction

•

Research problem

•

Prepare for viva

•

Objectives

•

Address reviewer feedback

•

Literature Review Section

•

Methodology Section

•

Results

•

Data analysis

•

Visualisations

•

Discussion

•

Interpret findings

•

Relate to objectives

•

Conclusion and
Recommendations

•

Summarise key points and
suggest future work

•

References using APA formatting

Using Napkin
Go to Napkin
Within this app you can then add the text you wish to use or generate using AI by adding a
prompt to generate text content. Once you have the text you wish to use, you can then
select all of it or a specific text section you wish to generate a visual for. This app provides
multiple options of visuals to choose from with options of different styles and background
colours. Each visual can then be downloaded as a PNG image and added to the
document of your choice.

81


Four stages of the academic writing process

Figure 1. Infographic with four stages of the academic writing process

Pre-writing
This involves topic selection, a literature review, and research design.

Figure 2. Infographic with the pre-writing process steps

82


Writing
This covers the creation of the introduction, literature review, methodology, results,
discussion, and conclusion sections.

Figure 3. Infographic with the writing process components

Post-writing
This focuses on proofreading, plagiarism checks, submission, and preparing for
evaluation.

83


Figure 4. Infographic with the post-writing process components and steps for successful submission

The next stage was to use NotebookLM to create an informative podcast. Using the
assessment brief and assessment criteria, alongside the stages of dissertation
infographic, these files were uploaded as PDFs.
NotebookLM then created a podcast discussing the content of the documents uploaded.
This highlighted the key components of academic writing structure, such as the
introduction, literature review, methodology, results, discussion, and conclusion. It also
emphasised the importance of keeping a research log, proper referencing and timely
submission.
Link to the: Dissertation Writing Stages podcast
The summary of the podcast created is:
The first source details the requirements and assessment criteria for a university
dissertation, including word count, formatting guidelines, ethical research
protocols, and marking scheme. It specifies the need for original research within
business or IT, emphasises ethical conduct, and warns against plagiarism,
especially using AI tools. The second source outlines the stages of dissertation
writing, from topic selection and literature review to data analysis, discussion, and
finally, submission and evaluation. Both sources aim to guide students through the
process of producing a high-quality dissertation.
This case study could be adapted in part or whole for any assessment brief.

84


•

Consider how visuals in the form of infographics could be used to capture key
points in an assessment brief or any user guides relating to the assessment.

•

Use NotebookLM to discuss and summarise an assessment brief and assessment
criteria. This can be used to supplement the written documents you will upload to
your VLE.

Links to Tools and Resources:
•

Dissertation Writing Stages podcast

•

Napkin – free infographic maker converting text to mind maps, flowcharts, and
more.

•

NotebookLM – used to create the podcast discussion

References:
Ritchie, J., Crooks, R. and Lankow, J. (2012) Infographics: The Power of Visual Storytelling.
Wiley.
Ware, C. (2020) Information Visualization: Perception for Design 4th edn. Morgan
Kaufmann.
Ware, C. (2021) Visual Thinking for Information Design 2nd edn. Morgan Kaufmann

85


Formative Assessment Activities for Using Generative AI
Responsibly and Ethically
Name: Sue Beckingham
Job Title and Affiliation: Associate Professor, Sheffield Hallam University
Contact Details: Sheffield Hallam University
Relevant Links:
Sheffield Hallam University - Sue Beckingham

LinkedIn - Sue Beckingham
What is the MML Idea?
The Generative AI CHECKLIST is a multimodal resource that can be used as a formative
assessment activity to discuss how this technology can be used responsibly and ethically.
It uses an infographic poster to outline nine points for consideration which can be used as
a handout. The infographic PDF was then used to create a NotebookLM podcast which
discusses what the checklist is and the benefits of using it. This audio recording can then
be shared with students as a summary of the checklist. To check understanding, ChatGPT
is used to create a multiple-choice quiz.
This example includes Generative AI elements of:
•

Image

•

Sound

•

Speech

•

Text

Why This Idea?
Research undertaken by Jisc (2023, 2024) with students has highlighted that there was a
lack of clear guidance from their institutions on how to use Generative AI (GenAI)
responsibly and ethically. The students wanted to understand what was considered
acceptable use, develop their information literacy, how to critically evaluate outputs and
also acquire GenAI skills that may be needed in future workplaces. We therefore have a
duty to help support students to do this by providing a collaborative sandpit to learn how
GenAI can be used appropriately (Beckingham et al, 2024).
The GenAI checklist was created to provide a multimodal formative assessment activity
where the tutor and students can engage in a discussion about the responsible use of the
growing number of tools available. The use of an infographic and podcast can be used to
structure the discussion and serve as a resource to revisit and check student
understanding. A low-stake formative test in the form of a quiz is then created to see what
the students recall and provide them with feedback (Mollick and Mollick, 2023).

86


The Generative AI CHECKLIST infographic outlines a checklist for responsibly using
generative AI. Key steps include validating AI-generated information with other sources,
formulating precise prompts to elicit desired results, and critically evaluating the AI's
responses for accuracy and bias. The checklist also emphasises the importance of setting
clear learning goals, expanding knowledge beyond the AI's output, and developing
effective learning habits that don't overly rely on generative AI tools. Finally, it advocates
for responsible usage, considering both the benefits and environmental impact.
The use of NotebookLM takes the checklist and creates a podcast summarising and
discussing each of the points listed. This provides users of the checklist a running
commentary bringing out the key messages. The audio recording is just 6 minutes, so it is
convenient to listen to.
This activity provides both a visual and audio resource to discuss the importance of
critically using generative AI. Finally, the multiple-choice quiz created using ChatGPT is
used as a formative low stakes test to gauge understanding.

How Could Others Implement This Idea?
You may wish to create your own checklist, guidance or a list of instructions. This is likely to
be initially either captured in a Word document or handwritten in a notebook.
•

Create an infographic to outline key points.
Go to Piktochart or an infographic maker of your choice. Select a template, colour
scheme and use the free searchable icons to reflect the information you are going
to add. Once completed you can save this as a PDF and an image file. The PDF can
be used for the next two resources; and the JPEG or PNG image file can be used to
embed in your module VLE and/or shared with students via social media.

•

Create an audio podcast as a summary
Go to NotebookLM and upload the PDF of the document you want to use. It will then
summarise the document in the form of a podcast.

•

Create a multiple-choice quiz to check students understanding
Go to ChatGPT and upload your PDF. Then adapt the following prompt which was
used for the checklist:
You are a quiz creator and I would like you to please give me a good low-stakes
test based on the Generative AI CHECKLIST [add your title]. I want to check
students' understanding of this infographic checklist [edit as appropriate] and set
them a minimum of 10 multiple choice questions. The questions should be highly
relevant and address each point on the checklist. Multiple choice questions should
include plausible, alternate responses and should not include an "all of the above
option." At the end of the quiz, you will provide an answer key and explain the right
answer.

87


Links to Tools and Resources:
•

Beckingham, S. and Hartley, P. (2024). The Generative AI CHECKLIST. National
Teaching Repository. Educational resource.

•

Beckingham, S. and Hartley, P. (2024). The Generative AI CHECKLIST. NotebookLM.

•

Canva – another free infographic maker

•

ChatGPT – used to create a multiple-choice quiz based on the content in the
infographic.

•

NotebookLM – used to create the podcast discussion

•

Piktochart – free infographic maker used to create the checklist which can then be
saved as a PDF and image file (JPEG or PNG)

References:
Beckingham, S., Lawrence, J., Powell, S. and Hartley, P. (2024) Using Generative AI Effectively
in Higher Education: Sustainable and Ethical Practices for Learning Teaching and
Assessment. Routledge. https://doi.org/10.4324/9781003482918
Jisc (2023) Student concerns around generative AI. Jisc National Centre for AI.
https://repository.jisc.ac.uk/9218/1/NCAI-Students-Perceptions-of-generative-AIReport.pdf
Jisc (2024) Student concerns around generative AI. Jisc National Centre for AI.
https://repository.jisc.ac.uk/9571/1/student-perceptions-of-generative-aireport.pdf
Mollick, Ethan R. and Mollick, Lilach, (2023) Using AI to Implement Effective Teaching
Strategies in Classrooms: Five Strategies, Including Prompts. The Wharton School
Research Paper. http://dx.doi.org/10.2139/ssrn.4391243

88


Live Storytelling through Generative AI Powered Avatars
Name: Simon Campion, Radoslaw Dorociak, Georg Meyer
Job Title and Affiliation: Virtual Engineering Centre, IDEAS, University of Liverpool
Contact Details: georg@liv.ac.uk
Relevant Links:
Virtual Engineering Centre

What is the MML Idea?
Our project aimed to develop an exhibit where learners could freely interact with a
historical character in a museum setting. The underlying technologies we used also apply
to any environment where users interact with virtual characters—past, present or future;
real or imagined.
The key components of our experience were:
•

A visualisation of the character in a historical setting that users could interact with,
implemented as an Epic MetaHuman character using historical photographs, as
well as facial and voice capture from an actor to create language models.

•

Text and speech interfaces in 40 languages, enabling users to ask questions in
their preferred language and modality.

•

An interface to a generative AI engine with predefined prior knowledge to ensure
responses were content-rich, appropriate, and historically accurate.

•

An adaptive database of common questions and answers, which helped reduce
synthesis time and enabled a more fluid user experience.

•

Text-to-speech, phoneme, and viseme synthesis modules to animate the
character. We matched the emotional valence of speech and expression to
context.

This example includes Generative AI elements of:
•

Image

•

Speech

•

Text

•

Video

Why This Idea?
Artificial intelligence-based storytelling can enhance learning by providing personalised,
interactive, and accessible education, which in turn improves engagement and outcomes
(Santally & Senteni, 2013). The use of experiential and self-paced learning also makes
interactive experiences highly accessible to diverse student populations (Leißau et al.,
2021).
89


Interactive AI-driven tools increase engagement by turning abstract concepts into
discussions, aligning well with constructivist learning theories (Vygotsky, 1978).
We first demonstrated the model at Liverpool’s St George’s Hall for the UK’s first Digital
Heritage Symposium in April 2022. Since then, the model has been used in an educational
tour during 2023 to celebrate Black History Month. During this tour, we supported children
and school groups as they engaged directly with the model, becoming immersed in their
learning experience.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
To replicate this idea, others could follow a similar process to ours:
1.

Create the character:
We visualised Mary Seacole using Epic MetaHuman. We based the digital model on
historical photographs and captured an actor’s facial movements and voice to
create the foundation for a believable digital character.

2.

Develop multimodal input and output:
We created both text and speech interfaces, supporting 40 languages. This made
it possible for users to interact in their preferred language and communication
mode.

3.

Integrate with generative AI and prior knowledge:
We connected the system to a generative AI engine, with predefined historical
knowledge to ensure accurate responses. We also built an adaptive Q&A database
to speed up interactions over time.

4. Animate responses dynamically:
Using text-to-speech, phoneme, and viseme synthesis, we gave the character the
ability to speak and display appropriate facial expressions. The emotional tone
was set based on context, enhancing realism.
5. Deploy in learning environments:
We launched the system in both public museum spaces and educational settings.
In each context, we supported learners in exploring the tool and evaluated their
interactions to inform improvements.
The approach is scalable and adaptable for any learning environment where interacting
with virtual personas could add educational value.

90


Links to Tools and Resources:
•

Video: Digital Humans | Making Mary Seacole

•

Case study: Reliving History: Bringing Historical Stories to Life through Digital
Avatars and AI Technology

•

Epic MetaHuman technologies

Image:

Figure 1. Avatar of Mary Seacole developed using Epic MetaHuman technology, designed for realtime interaction with learners in museum and educational settings

References:
Leißau, M., Hellbach, S., Laroque, C., Busch, C., Steinicke, M., Friess, R., & Wendler, T. (2021,
October). Self-paced learning in virtual worlds: Opportunities of an immersive
learning environment. In Proc. Eur. Conf. e-Learn., ECEL (pp. 257-265). Academic
Conferences and Publishing International Limited.
Santally, M. I., & Senteni, A. (2013). Effectiveness of Personalised Learning Paths on Students
Learning Experiences in an e-Learning Environment. European Journal of Open,
Distance and E-learning, 16(1), 36-52.
Vygotsky, L.S., 1978. Mind in society: Development of higher psychological processes. In: M.
Cole, V. John-Steiner, S. Scribner and E. Souberman, eds. Cambridge, MA: Harvard
University Press. Available at: https://doi.org/10.2307/j.ctvjf9vz4
91


Strategies for Integrating AI-generated (GenAI) Technologies in
Assessment
Name: Mari Cruz García Vallejo
Job Title and Affiliation: Digital Education Consultant and Affiliated Lecturer at the ULPGC
(Spain)
Contact Details: maricruzgarciavallejo@gmail.com
Relevant Links:
AI Pedagogy Project - Mari Cruz Garcia Vallejo

What is the MML Idea?
This bilingual case study presents a practical design guide intended for both novice and
experienced higher education lecturers. It provides strategies for integrating AI-generated
(GenAI) technologies in assessment within their teaching practices. The design guide is
based on the concept of authentic assessment, a practical approach that enables
students to connect and apply knowledge gained in the classroom to real-world tasks or
projects relevant to their academic or professional disciplines. Additionally, the guide
promotes assessment for learning, particularly through the modality of assessment as
learning, by shifting the assessment focus to students themselves, who take on the roles
of assessors of their own work and that of their peers.
From a theoretical standpoint, the design guide builds on the assessment models
proposed by Villarroel et al. (2018) and Sambell (2022), enhancing and expanding these
frameworks with an additional layer on “how to integrate GenAI in assessment.”
The guide begins with the premise that AI integration in assessment within higher
education can be addressed through three distinct approaches:
•

Assessment of AI

•

Assessment with AI

•

Assessment for AI

(The guide offers further detail on each of these three approaches.)
The guide is organised into four design steps to assist lecturers in developing assessment
methods tailored to their courses:
1.

Contextual Analysis.

2.

Designing the Assessment Task, encompassing technology selection and the
chosen approach for AI integration.

3.

Developing Marking Criteria, including the creation of rubrics and assessment
standards.

4. Feedback Process.
92


Each design step equips lecturers with reflective questions, analyses of pros and cons,
examples of well-designed assignments, and resources to guide them through the
assignment design process. The four steps enable lecturers to assess and determine the
role of students within the assessment process, taking into account factors such as
student demographics, familiarity with technology and GenAI tools, and their level of
assessment literacy.
The case study concludes with recommendations for the ethical use and integration of
GenAI in assessment, drawing on feedback from participants in a final course survey.
This example includes Generative AI elements of:
•

Text

Why This Idea?
The integration of Generative AI (GenAI) into teaching and assessment is increasingly vital
for higher education (HE) across Spain, the EU and the UK. With GenAI technology
advancing at a rapid pace, lecturers need a structured and strategic approach to
incorporating GenAI in ways that extend beyond traditional “assessment of learning”
models.
This design guide was created as a practical toolkit within the framework of the ULPGC
module “CAIE24 Desarrollo de competencias en IA aplicadas a la evaluación: hacia una
evaluación más inteligente” (Development of AI Competencies Applied to Evaluation:
Towards a More Authentic Assessment). Part of the “Plan de Formación 2024-2025” at the
Universidad de Las Palmas de Gran Canaria (ULPGC), this 1 ECTS postgraduate-level
module targets ULPGC teaching staff is -roughly-equivalent to PGCAP/PGCert
programmes in HE in the UK. It introduces participants to AI competency frameworks and
AI literacy, equipping them with strategies for effective GenAI integration in assessment
practices. This initiative supports the development of AI-related skills, knowledge, and
competencies as outlined in both EU and Spain´s national frameworks for AI competencies
in HE.
The guide provides a structured, theoretically grounded resource that helps lecturers new
to assessment understand the purpose and role of GenAI in contemporary education.
Beyond the practical benefits, the guide supports lecturers in reflecting on and
reconceptualising current assessment practices to cultivate competencies crucial for new
employability attributes. Endorsed by the Vicerrectorado de Innovación Educativa at
ULPGC, this initiative provides a valuable resource for educators striving to integrate GenAI
ethically and effectively, with preliminary applications indicating positive impacts. It
stands as a scalable model for transformative assessment practices in HE institutions
across the EU.

93


How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
The four steps of the guide:
1.

Context.

2.

Designing the assessment task, which includes the choice of technology and the
approach for the integration of the IA.

3.

Designing how to mark the task, which includes the design of rubrics and
assessment criteria.

4. Process of feedback.
This can be adapted to different contexts and academic disciplines.
This is a graphic representation (in Spanish) of how to adapt the four steps. The graphic
can be adapted into other languages.

Figure 1. A graphic representation of adapting the four steps of the guide

Links to Tools and Resources:
The graphic provided in the previous section will be translated into English for the final
case study.

References:
Sambell, K. (2022). How to design assessments for learning. Heriot-Watt University.
https://lta.hw.ac.uk/wp-content/uploads/22_AFL_How-to-design-assessmentsfor-learning.pdf
Villarroel, V., Bloxham, S., Bruna, D., Bruna, C., & Herrera-Seda, C. (2018). Authentic
assessment: Creating a blueprint for course design. Assessment & Evaluation in
Higher Education, 43(5), 840–854. https://doi.org/10.1080/02602938.2017.1412396

94


Experimental Practice – Using Mind Maps in Project Planning to
Foster Inclusion
Name: Gareth Ellis
Job Title and Affiliation: University Teacher, University of Liverpool
Contact Details: gdellis@liverpool.ac.uk

What is the MML Idea?
This approach combines AI tools with mind mapping to enhance groupwork skills through
both text-based and visual learning methods. I use GenAI to support structured project
planning, where students generate task lists and responsibilities using large language
models (LLMs), then transform these into interactive mind maps. This multimodal strategy
helps visualise elements and stages of complex group projects.
This example includes Generative AI elements of:
•

Image

•

Speech

•

Text

Why This Idea?
Combining AI tools with mind mapping offers a unique way to enhance groupwork skills
through both text-based and visual learning methods. Seminal theoretical frameworks in
cognitive load theory suggest that breaking tasks into smaller, structured components
reduces mental effort, enabling deeper learning (Sweller, 1988). The integration of AI and
mind mapping helps students manage complex group projects by visualising elements
and timelines more effectively.
Research on AI in education demonstrates its ability to boost productivity, critical thinking,
and organizational skills (Essien et al., 2024). Similarly, studies on mind mapping show its
effectiveness in supporting comprehension and fostering creative problem-solving
(Erdem, 2017). By merging these tools, I aim to leverage the best of both, offering students
a structured yet flexible way to approach group tasks.
We collected student feedback from my support sessions which highlighted that students
feel more engaged and better prepared when using this method. The feedback also
suggested that the process was intuitive, enabling students to quickly adopt the method
with minimal technical instruction.
Students often struggle with managing group projects due to unclear task allocation,
uneven participation; and unfocused, asynchronous communication. This approach
addresses these challenges by creating a structured, visually accessible plan, which can
95


be easily explained to and shared amongst group members The use of AI-generated
Markdown language facilitates accessibility for students with diverse profiles, supporting
inclusion through logical, visual task breakdown. This method supports development of
transferable skills in project management, digital literacy, and collaborative problemsolving—key competencies for academic and professional success.

How Could Others Implement This Idea?
1.

Introducing the Assessment Brief
Students are taken through the assessment brief in detail.
A plenary session follows for resolving doubts and clarifying expectations.

2. Customising the Prompt for the AI Tool
Students are introduced to the following prompt template:
"We are a group of students working on a project for a [Course Title] course at
[Course Level] in a [Context] university, titled [Project Title]. Our goal is to [Describe
Project Goal]. We need help creating a project plan that includes:
•

Key tasks and subtasks

•

Roles and responsibilities for each task

•

Timeline with deadlines

•

Milestones and check-in points

•

Expected deliverables

Please generate this plan in both Markdown format and professionally formatted
DOCX format. Make the Markdown format compatible with and suitable for the
creation of a mind-map."
3. Generating the Plan
Students access the LLM of their choice (free or paid).
They input the customized prompt either individually or in groups.
They retrieve both Markdown and DOCX outputs from the LLM.
4. Visualising the Plan with Markmap.js
The tutor demonstrates copying the Markdown output into Markmap.js.
They generate a visual mind-map, showing how to expand, collapse, and edit
sections.
5. Student Practice
Students copy their visual Markdown output into Markmap.js.
They view and refine the generated mind maps collaboratively.
Once finalized, they save the mind map as an interactive HTML page for project
documentation.
6. Presentation and Reflection (Optional)
96


If time allows, each group presents their mind map.
Presentations include:
•

Project goals

•

Key tasks and assigned roles

•

Challenges faced and solutions developed during planning

Links to Tools and Resources:
•

OpenAI-ChatGPT

•

Anthropic-Claude

•

Google-Gemini

•

gera2ld-Markmap

References:
Erdem, A. (2017). Mind maps as a lifelong learning tool. Universal Journal of Educational
Research, 5(12A), 1–7. https://doi.org/10.13189/ujer.2017.051301
Essien, A., Bukoye, O., O’Dea, C., & Kremantzis, M. (2024). The influence of AI text generators
on critical thinking skills in UK business schools. Studies in Higher Education, 49, 865
- 882. https://doi.org/10.1080/03075079.2024.2316881.
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive
Science, 12(2), 257–285. https://doi.org/10.1207/s15516709cog1202_4

97


Enhancing Learning with AI Tools for Idea Generation
Name: Dr Katherine Geer, Daniel Hsu
Job Title and Affiliation:
Lecturer in Marketing, Liverpool John Moores University,
Lecturer in Digital Marketing & Analytics, Liverpool John Moores University
Contact Details: K.A.Geer@ljmu.ac.uk, T.Hsu@ljmu.ac.uk

What is the MML Idea?
Using a design thinking approach to generate ideas (Bustard et al., 2022; Seevaratnam et
al., 2023), we applied multimodal teaching using AI tools to support a cohort of 198 level 5
undergraduate students undertaking a Media Production Management module. Students
were tasked with designing three pieces of content for three different digital platforms,
including video and podcast formats, for their real-world client.
By applying text, image and audio tools, we aimed to empower students with varied
pathways to understand, explore and implement ideas for content creation. Leveraging
ChatGPT, MS Copilot, and NotebookLM, students engaged in an iterative process to
develop their creative thinking skills, while being encouraged to critically assess AI outputs
throughout (Huang, 2023). This case study demonstrates how multimodal AI enhanced
learning across three stages.
This example includes Generative AI elements of:
•

Image

•

Sound

•

Text

Why This Idea?
We wanted students to experience a structured, creative process for content generation
that mimicked real-world workplace scenarios. Applying the design thinking stages of
Understand, Explore, and Materialise, we guided students in using GenAI tools as support
systems, not shortcuts.
Firstly, students used ChatGPT to research ideas that aligned with their client’s brief and
would appeal to their target audience. Students were guided to use text to create openended prompts such as “inclusive marketing campaign examples for students” or
“community focused content ideas for Gen Z.” ChatGPT provided instant responses,
offering diverse ideas and relevant examples for students to critically assess, before
defining potential content ideas.
Once students had defined their content idea, MS Copilot was introduced to facilitate
visual exploration. Through Copilot, students generated images to transform abstract
98


ideas into tangible visuals. Students used their AI-generated images to craft mood boards
and storyboards to communicate a prototype of their content idea to the client. This stage
fostered creativity through exploration, empowering students to push beyond text-based
ideas and refine their vision through visual expression and criticality.
In the final stage, students used NotebookLM to generate audio narrations for their
podcast. Use of NotebookLM enabled students to transform their brainstorming activities
and critical thinking into professional sounding audio. Through listening to each other's
audio responses, students gained exposure to different AI outputs within their team, and
critically reflected on how the more successful outputs had been achieved. This
collaborative exchange enabled them to refine final content in a way that met client
expectations and resonated with the target audience.
Students gained current workplace design skills by learning how AI can facilitate idea
generation (see Figure 1). They also learnt to think critically about AI outputs, an essential
skill for successful application of AI in the workplace.

How Could Others Implement This Idea?
Below are the steps we followed:
1.

Stage 1: Understand
Guide students to use ChatGPT with open-ended prompts based on a client brief
or topic. Encourage them to refine outputs through critical questioning. Prompts
could include:
•

“Create content ideas for Gen Z that focus on community-building.”

•

“Suggest marketing campaign themes for a student-focused brand.”
This stage supports early idea generation and critical thinking.

2.

Stage 2: Explore
Ask students to take one chosen idea and use MS Copilot (or another imagegeneration tool) to visualise it. They can:
•

Generate images to represent mood or message

•

Assemble mood boards or storyboards
These visuals help clarify and pitch the concept.

3.

Stage 3: Materialise
Students use NotebookLM (or a similar AI narration tool) to transform their ideas
into script and audio form. They can:
•

Develop podcast narration based on their storyboard and research

•

Share and critique each other’s outputs
This encourages reflection, collaboration and iterative refinement.

99


All tools used—ChatGPT, MS Copilot, and NotebookLM—were accessed in their free versions
during this activity. The process is adaptable across disciplines where creative or
campaign-based outputs are required.

Image:

Figure 1. Example of student work (storyboard outcome)

References:
Seevaratnam, V., Gannaway, D., & Lodge, J. (2023). Design thinking-learning and lifelong
learning for employability in the 21st century. Journal of Teaching and Learning for
Graduate Employability, 14(1), 167–186.
Bustard, J. R. T., Hsu, D. H., & Fergie, R. (2022). Design thinking innovation within the
quadruple helix approach: a proposed framework to enhance student
engagement through active learning in digital marketing pedagogy. Journal of the
Knowledge Economy. https://doi.org/10.1007/s13132-022-00984-1
Huang, C.-W., Coleman, M., Gachago, D., & Jean-Paul Van Belle. (2023). Using ChatGPT to
encourage critical ai literacy skills and for assessment in higher education.
Communications in Computer and Information Science, 105–118.

100


Enhancing Engagement through Streaming Avatars and
Synchronous AI Assistants in Online Education
Name: Dr Ioannis Glinavos
Job Title and Affiliation: Senior Lecturer in Law, University of Westminster
Contact Details: i.glinavos@westminster.ac.uk
Relevant Links:
University of Westminster - Dr Ioannis Glinavos

What is the MML Idea?
The proposed case study explores how streaming avatars and synchronous AI assistants
can enhance engagement in online education. I created the avatars that were used for a
dynamic experience and AI assistants for real-time support, and this approach fosters an
interactive learning environment. The study will assess the impact on learner satisfaction
and provide insights into effective integration practices.
This example includes Generative AI elements of:
•

Image

•

Sound

•

Space

•

Speech

•

Text

•

Video

Why This Idea?
The integration of streaming avatars and synchronous AI assistants in online education
addresses the growing need for more engaging and personalized learning environments.
Theoretically, these technologies enhance social presence and reduce transactional
distance, promoting a stronger sense of community among learners. Practically, avatars
and AI assistants provide an adaptable, inclusive way to deliver content while mitigating
privacy concerns and presentation fatigue. Empirical evidence suggests that these tools
can improve learner engagement and satisfaction by offering just-in-time support and
increasing interaction. This idea meets the pressing need for innovative, evidence-based
approaches that foster student engagement and well-being in digital education.

How Could Others Implement This Idea?
The integration of streaming avatars and synchronous AI assistants in online education
addresses the need for engaging and personalized learning environments by enhancing
social presence, reducing transactional distance, and promoting community. These
technologies provide adaptable ways to deliver content, mitigating privacy concerns and
101


presentation fatigue, and empirical evidence suggests they improve learner engagement
and satisfaction. Educators can implement this idea by selecting suitable platforms and
tools, developing tailored prompts, testing the setup, and adapting based on student
feedback to foster an interactive and inclusive online learning experience.

Links to Tools and Resources:
•

Example of a streaming avatar in class

•

Example of an AI assistant capable of being embedded in a VLE

References:
Anderson, T. (2003). Getting the mix right again: An updated and theoretical rationale for
interaction. The International Review of Research in Open and Distributed Learning,
4(2). https://www.irrodl.org/index.php/irrodl/article/view/149
Clark, R. C., & Mayer, R. E. (2016). E-learning and the science of instruction: Proven
guidelines for consumers and designers of multimedia learning. Wiley.
https://onlinelibrary.wiley.com/doi/book/10.1002/9781119239086
Hrastinski, S. (2008). Asynchronous and synchronous e-learning. Educause Quarterly,
31(4), 51-55.
https://www.researchgate.net/publication/238767486_Asynchronous_and_synch
ronous_e-learning

102


How much can AI (at present) Contribute to Identifying and
Writing about Classification and Nature?
Name: Tom Hartman
Job Title and Affiliation: Lecturer in Biology, University of Nottingham
Contact Details: thomas.hartman@nottingham.ac.uk
Relevant Links:
LinkedIn - Tom Hartman

What is the MML Idea?
I developed a rubric which could be used for student assessment. Students were provided
with various elements generated about birds and nature, and then they were asked to
give a critical evaluation of the AI derived essays produced from these elements.
This example includes Generative AI elements of:
•

Image

•

Object

•

Sound

•

Speech

•

Text

•

Video

Why This Idea?
One of the key issues that conservation organisations deal with is the discernible lack of
skills with which biology graduates have in identifying organisms in the wild. One of the
aims of the MSc in Biological Photography and Imaging is to provide a basic grounding in
natural history so that the students may not only take pin sharp and accurate
photographs of a living organism or biological specimen, but also be able to identify it
accurately. This has been achieved through training on using dichotomous keys and other
guides, but recent developments have offered many new techniques. To supplement their
training, students were given briefs to identify birdsong using the Merlin app, take images
of different species and identify them using Google Lens and then critically evaluate text
on the locality and wildlife written by ChatGPT. They were then asked to mark the AI
derived essay and write a reflection on the experience. Although this is a preliminary study
with low data (n=10) the universal themes that were expressed were that the AI generated
text reads authentically and is well expressed. It was only when the text was probed more
deeply that the flawed arguments, fictitious referencing, unlikely observations and vague
descriptions became noticeable. It does score well in the breadth of scope and this was
well received by this group of students. Being asked to critique an AI derived essay had a

103


considerable effect on their perception of the authority of generative AI text, its research
parameters and how to phrase their prompts.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
Follow the rubric that I developed and examine the responses using sentiment analysis.
An example can be found below:

Course work on artificial intelligence and Natural History.
The BIRD.
Teaching Aims: To investigate the way in which modern technology can enhance our
understanding of the natural world and whether there are limitations and restrictions that
they impose upon developing the skills to do things in the wild.
Objective 1: Choose an area of the campus and sample the bird song using the Merlin
app (from the App store) to work out what birds frequent this area at three different times
of the day. Create a list of the birds that you hear at different times of the day and
produce a graph of the number of species singing throughout the period of your
evaluation. You will need to keep your device running for 15 minutes or so.
Objective 2: Photograph a number of the birds and identify them using Google Lens (from
the App store) or some similar programme.
Objective 3: Use ChatGPT or Gemini to develop a 400 word paragraph to describe the
area and the birds that you have identified using Merlin and Google lens. Ask it to include
references.
You could even attempt to produce an opening image using image generating
programmes such as DALL·E

Assessment:
1.

Add the bird information and images to your other four organisms.

2.

Mark the AI generated text with a few statements giving some feedback (a few
words or a sentence). Look for accuracy and use of language. Check its references

3.

Rewrite and augment the ChatGPT or Bard AI generated text and, using your
images, develop a short article on this area of the university for inclusion in your
website. Include the number of species revealed by using the Merlin app.

Assessment: Write a 500 word personal reflection about what it was like using these
programmes and whether they will be a help or hindrance to people developing skills in
Natural History and identification. This should be written as a description of your personal
104


experience, thoughts and understanding of your encounter and or partnership with
artificial intelligence.
•

Do you think that this is an asset or will be detrimental to developing field skills and
field craft?

•

How do you feel about technology supplementing or replacing human creativity or
should we consider this to be an aid such as using a calculator or grammar
checker?

•

What sources are these AI programs using?

•

Is this plagiarism?

•

Is this an extra barrier between the original information and the user?

Marking criteria
•

Selection of key organisms and criteria for diagnosing the species

•

Image quality, layout and writing

•

Taxonomy and species description

•

Reflective commentary

Reports will be marked for:
•

Photography

•

Layout

•

Correct ID

•

Positive ID reasoning

•

Writing and reasoning and reflection.

Links to Tools and Resources:
•

Gemini

•

ChatGPT

•

DALL·E

105


Using Generative AI to Aid Studying and Learning for Medical
Students
Name: Ayaat Jaway
Job Title and Affiliation: Student Advisor, University of Liverpool
Contact Details: hlajaway@liverpool.ac.uk

What is the MML Idea?
I’ve found that using ChatGPT can support medical students in studying complex content
by combining image and text-based learning. Generative AI is particularly helpful for
breaking down difficult concepts, providing clear revision summaries, and even
generating diagrams to reinforce understanding. By using multimodal prompts—both text
and image—I can guide students through topics like the RAAS system in a way that builds
confidence and comprehension.
This example includes Generative AI elements of:
•

Image

•

Text

Why This Idea?
It can often be difficult to study content provided by lecturers, and some topics are more
challenging to understand, making revision an obstacle. Medicine requires consistent
studying and I’ve seen how online tools like ChatGPT (OpenAI) can be helpful for breaking
down difficult subjects and preparing for exams. In addition, the process of “scaffolding” in
learning (Wood et al., 1976) can be easily applied by adapting prompts in ChatGPT to
explain concepts at a basic level, and then deepening the level of detail in follow-up
prompts to support progression.
ChatGPT is already being used widely by university students, and in one study it was
assessed on university-level biochemistry exams (Mahat et al., 2023). It was even reported
to outperform medical students in a physiology exam (Soulage et al., 2024). Now that
ChatGPT is multimodal, it can also generate visual outputs, such as diagrams. This feature
adds significant value for visual learners and supports more engaging and accessible
revision materials.

How Could Others Implement This Idea?
1.

Firstly, attach any notes taken to ChatGPT. This makes it easier for ChatGPT to
generate relevant outputs at the right level. The attaching note’s function should
be free and can be accessed with an account. Alternatively, typing the name of
the topic in the ChatGPT search bar could generate notes although these would
need to be checked for accuracy before proceeding.
106


2.

Use a prompt to help with studying the topic related to the notes taken. Here is an
example prompt:
‘Summarise the RAAS system in a way that is easy to understand and in no
more than 200 words.’

3.

Then, to accompany the text generated, use a prompt to generate a relevant
image:
Provide an image with a diagram representing the RAAS system, at the level
of knowledge required in a medical school exam. Use the text summary
generated to label the diagram.

4. The prompt can also be adapted to provide exam questions for practice:
Summarise the RAAS system in a way that is easy to understand and in no
more than 200 words. Then create 5 challenging multiple-choice questions
in a single best answer format so I can test my understanding of the topic.

Links to Tools and Resources:
•

Open AI (2024)

Image:

Figure 1: RAAS system diagram

Note: The diagram was produced from ChatGPT (OpenAI) using the prompts above.
OpenAI. (2025). RAAS system diagram [AI-generated image]. ChatGPT.

107


References:
Mahat, R. K., Jantikar, A. M., Rathore, V., & Panda, S. (2023). Assessing the performance of
ChatGPT to solve biochemistry question papers of university examination.
Advances in Physiology Education, 47(3), 528–529.
https://doi.org/10.1152/advan.00076.2023
Soulage, C. O., Fabien Van Coppenolle, & Fitsum Guebre-Egziabher. (2024). The
conversational AI “ChatGPT” outperforms medical students on a physiology
university examination. AJP Advances in Physiology Education.
https://doi.org/10.1152/advan.00181.2023
Wood, D., Bruner, J. S., & Ross, G. (1976). The Role of Tutoring in Problem Solving. Journal of
Child Psychology and Psychiatry, 17(2), 89–100.
OpenAI. (2024). ChatGPT (GPT-4o, May 13 version) [Large language model].
https://chat.openai.com/chat

108


Exploratory Learning using GAI to Support Presentations
Name: Liam Kaye and Chris Barlow
Job Title and Affiliation: ULMS Liaison Librarian; Senior Lecturer Accounting and Finance
Contact Details: liam.kaye@liverpool.ac.uk; chbarlow@liverpool.ac.uk
Relevant Links:
Library - Liam Kaye
University of Liverpool - Chris Barlow

What is the MML Idea?
We introduced first-year Accounting and Finance students to Generative Artificial
Intelligence (GAI) through an exploratory task with minimal tutor intervention. The idea
was to allow students to engage directly with different GenAI tools, observe how they work,
and assess their strengths and limitations within a real-world academic task.
We delivered the session through a short lecture, an interactive Kahoot quiz, group-based
presentation development, and verbal feedback discussions. Students used GenAI
platforms (ChatGPT, CoPilot, Gemini, and Perplexity) to research company sustainability
initiatives and prepare a group presentation. Each group was also asked to evaluate their
assigned tool. Modes of communication used in the session included speech, text, and
group interaction.
This example includes Generative AI elements of:
•

Image

•

Speech

•

Text

Why This Idea?
Generative Artificial Intelligence (GAI) is of great interest to information professionals in HE,
particularly when discussing their potential impacts regarding academic integrity and
information literacy (Ilieva et al, 2023). While there’s often an assumption that students
are already using these tools confidently, research shows that student familiarity and
understanding vary significantly (Madunić, 2024). Many students still have only a basic
understanding of what GenAI is, what tools exist, how they function, and how they can
support academic work. We designed this session with these gaps in mind. We also
wanted to explore evidence suggesting that less guided, exploratory approaches—where
students manage their own error correction—can be highly beneficial in building an
understanding of GenAI and large language models (Barrett & Stout, 2024).
We therefore constructed a session that introduced an overview of GAI to our students,
before allowing them to explore GAI’s capabilities through a given task. Students would
109


see first-hand how GAI functions and where its strengths could be applied to the module
(Skills for the Professional Accountant), which had a focus on skills development, including
relevant resource usage, presentation, and communication skills.
We first introduced the general concepts of GAI using an Introduction to GAI Kahoot quiz
(2024), giving students a basic understanding underpinning how GAI works. This helped
equip them with the knowledge to approach the tasks with more confidence, as we
wanted them to explore the capability of GAI with as little guidance as possible. The
Kahoot covered what GAI are, their development, their reliability in different situations, how
their information is updated, and whether or not they can be used in the University.
Once this base understanding was covered, we split the classes into four groups and
directed each to a different GAI (ChatGPT, CoPilot, Gemini, or Perplexity) and a different
company to focus on (Coca-Cola, Apple, Microsoft, or Starbucks). They were then given 40
minutes and told to use their GAI to help them create a 5-minute presentation answering
the question:
“What are your companies' main sustainability initiatives and how impactful have they
been to the business and society?”
They were also asked to include a short review on the strengths and weaknesses of their
GAI. This would then allow for discussion among the group.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
1.

Give an introductory overview to ensure students have a basic understanding of
GAI

2.

Split the class into groups and assign each a GAI

3.

Assign a task all teams must complete using their GAI (tailor this to fit the learning
outcomes of the module)

4. Have each group feedback to the whole class, providing their assessment of the
GAI and open for questions/ discussion
5. End by summarising the strengths and weaknesses of GAI in regards to this
module specifically. Also take time here to error control, clear up any obvious
misunderstandings.

Links to Tools and Resources:
•

Kaye, L. (2024) Introduction to GAI. Kahoot.

•

ChatGPT (Free version) - ChatGPT

•

Copilot - Microsoft Copilot

•

Google Gemini - Gemini

•

Perplexity - Perplexity
110


Image:

Figure 1. Example of Student-created visual from a group presentation exploring Coca-Cola’s
sustainability initiatives using Generative AI tools.

References:
Barrett, L. and Stout, D. (2024). Minds in movement: embodied cognition in the age of
artificial intelligence. Philosophical Transactions of the Royal Society B: Biological
Sciences, 379(1911). https://doi.org/10.1098/rstb.2023.0144
Ilieva, G, et al. (2023). Effects of Generative Chatbots in Higher Education. Information,
14(9), 492. https://doi.org/10.3390/info14090492
Madunić, J., & Sovulj, M. (2024). Application of ChatGPT in Information Literacy Instructional
Design. Publications, 12(2), 11. https://doi.org/10.3390/publications12020011

111


Peer Conflict Resolution in Group Work: Using GenAI to Represent
Subject Knowledge Multimodally
Name: Rob Lindsay
Job Title and Affiliation: Educational Developer, University of Liverpool
Contact Details: rob.lindsay@liverpool.ac.uk

What is the MML Idea?
I created an activity which is the Peer Conflict Resolution activity and it is an interactive
resource designed to help students navigate common conflicts in university group
projects. This resource uses multimodal GenAI elements, including avatar videos,
synthesised voices, and reflective prompts, to create realistic conflict scenarios. In each
scenario, students are presented with options, receive feedback on their choices, and are
provided with the opportunity for reflection.
This example includes Generative AI elements of:
•

Image

•

Speech

•

Text

•

Video

Why This Idea?
This activity addresses the frequent challenge of managing interpersonal conflict in group
work, which is essential for academic and professional collaboration. By immersing
students in realistic scenarios with immediate feedback, the tool encourages selfreflection on communication and conflict resolution strategies, promoting cultural
competence (De Vita, Carroll, & Ryan, 2005; Asgari, 2019). Research indicates that effective
conflict resolution is crucial for positive team dynamics and outcomes, especially when
managing the varied types of conflict that can arise in team environments (Behfar,
Peterson, Mannix, & Trochim, 2008). The use of multimodal, interactive elements in this
tool—such as avatar videos and synthesized voices—aligns with established methods that
have proven effective for assessing and developing conflict resolution skills in realistic
settings (Olson-Buchanan et al., 1998).

How Could Others Implement This Idea?
Scenario Design
•

Task: Use Generative AI to create scenarios about group work conflicts and refine
them based on feedback.

•

Prompt Example: "Create a set of brief group project scenarios (3-5 sentences
each) that feature common issues such as unequal contributions,
112


miscommunication, clashing personalities, and scheduling conflicts. In each
scenario, ensure diversity of perspective and have one team member express their
frustration or thoughts about the conflict in a short transcript. For each scenario,
generate three possible responses as multiple-choice options, where each choice
represents a different way to address the problem. Include constructive feedback
for each choice, explaining why it’s effective or not in resolving the conflict."
•

Refinement Prompt: "Modify this scenario to focus more on cultural
communication breakdowns."

•

Output: Export all final scenarios and resources into one document for easy
access.

Audio Creation
•

Task: Use text-to-speech to upload each transcript and produce character voices.

•

Prompt Example for Voice Selection: "Create a calm, assertive female voice for the
team leader character."

•

Export: Download each audio file as an mp3 for future steps.

Create Video Avatars
•

Task: Use Adobe Express Avatar Creator to create avatars.

•

Process: Upload the generated audio files and choose an avatar that represents a
university student, with a suitable background.

•

Export: Download each video avatar.

Interactive Setup in H5P
•

Task: Embed scenarios in your virtual learning environment (VLE) or use H5P. Add
question sets and reflection prompts.

Feedback Survey Creation
•

Task: Use Generative AI to create relevant feedback survey questions, adjusting
these based on the project’s focus.

•

Prompt Example: ‘Consider this interactive resource. Create a feedback survey for
students, asking how effective they found the scenarios and what improvements
they suggest.’

To adapt this concept to your academic discipline, adjust the scenario and dialogue
prompt in the design stage to reflect any relevant challenges. Whether in healthcare,
business, education, or another discipline, swap in typical issues like communication,
decision-making, or policy dilemmas. This small change makes the learning experience
practical and transferable, keeping the core structure while tailoring content to your
teaching and students’ learning needs.

113


Links to Tools and Resources:
•

Peer conflict resolution activity – hosted on the H5P platform and can be
embedded.

•

ChatGPT – for refining scenarios, dialogue and feedback (free plan available)

•

ElevenLabs – GenAI text to speech with customisable voices (free plan available)

•

AdobeExpress – Free avatar creator with audio upload

•

H5P.com - Interactive resources platform for hosting (requires licence)

•

University VLE – for alternative hosting to H5P e.g. Canvas or Blackboard (free
institutional access)

References:
Asgari, S. T. (2019, July). Multi-cultural group work. University of Liverpool, Centre for
Innovation in Education. Retrieved from
https://www.liverpool.ac.uk/media/livacuk/centre-for-innovation-ineducation/staff-guides/multi-cultural-group-work/multi-cultural-group-work.pdf
Behfar, K., Peterson, R., Mannix, E., & Trochim, W. (2008). The critical role of conflict resolution
in teams: A close look at the links between conflict type, conflict management
strategies, and team outcomes. Journal of Applied Psychology, 93(1), 170–188.
https://doi.org/10.1037/0021-9010.93.1.170
De Vita, G., Carroll, J., & Ryan, J. (2005). Fostering intercultural learning through
multicultural group work. Teaching International Students: Improving Learning for
All, 75–83.
Olson-Buchanan, J., Drasgow, F., Moberg, P., Mead, A., Keenan, P., & Donovan, M. A. (1998).
Interactive video assessment of conflict resolution skills. Personnel Psychology,
51(1), 1–24. https://doi.org/10.1111/j.1744-6570.1998.tb00714.x

114


Image:

Figure 1. The peer conflict resolution activity landing page

Note. The peer conflict resolution activity landing page with introductory text “Navigate
common group project challenges with this interactive activity, designed to enhance your
conflict resolution and teamwork skills through real-life scenarios and guided feedback.”

115


A Multimodal Groupwork Activity: Fostering Belonging Among
First-Year Bioscience Students
Name: Dr Alice Maher
Job Title and Affiliation: Lecturer, University of Liverpool
Contact Details: Alice.Maher@liverpool.ac.uk
Relevant Links:
University of Liverpool - Dr Alice Maher

What is the MML Idea?
We asked students to work within their tutorial group to design an organism and present it
collaboratively. Each group was given three future Earth scenarios, all grounded in the
context of the climate emergency. Their task was to imagine what kind of organism—
plant, animal, or bacteria—might evolve under those conditions, while staying within
realistic physical and evolutionary constraints.
To bring their organisms to life, students were encouraged to use a range of creative
outputs including drawing, painting, physical models, and generative AI tools. By doing so,
they explored how multimodal tools can help represent biological adaptation and
speculative evolution in a future Earth setting.
This example includes Generative AI elements of:
•

Image

•

Text

Why This Idea?
We designed this activity to meet several learning objectives:
•

Encourage students to think critically and creatively about how organisms might
evolve under different climate-change scenarios, grounded in biological reality.

•

Provide experience in working effectively in groups, contributing to shared outputs
and communication.

•

Give students the opportunity to practise delivering group presentations.

•

Introduce students to generative AI as a tool for creatively visualising their ideas
and supporting multimodal learning.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
1.

Scenario Design:
Present students with a selection of future Earth scenarios shaped by climate
116


change. These could vary in temperature, rainfall, or atmospheric conditions.
Students choose one and begin brainstorming.
2.

Organism Development:
In groups, students design an organism that might plausibly evolve under their
chosen conditions. They research current organisms living in similar environments,
identifying useful biological adaptations.

3.

Creative Output:
Students create a visual representation of their organism using any medium—
hand drawing, painting, 3D models, or generative AI image tools. They must include
their organism in the final presentation and explain how each adaptation helps it
survive in the scenario.

4. Presentation:
Groups present their work in an 8-minute talk. Every member is required to speak.
The presentation includes:
•

Background: Environmental context based on the chosen scenario.

•

Inspiration from existing organisms: What adaptations already exist that
inspired their design.

•

Final design: A walkthrough of their organism’s features, based on
evolutionary logic.

Image:

117


Figures 1, 2 and 3. Examples of GAI created organisms from the 2024-25 year 1 BIOS105
communication skills module.

118


AI Visual Metaphor Activity (SEDA Learning to Tutor Online
Course)
Name: Elora Marston and Jonathan Rhodes
Job Title and Affiliation: Academic Developers, The University of Wolverhampton
Contact Details: e.marston@wlv.ac.uk and j.rhodes2@wlv.ac.uk
Relevant Links:
LinkedIn - Elora Marston
LinkedIn - Jonathan Rhodes
University of Wolverhampton - Jonathan Rhodes

What is the MML Idea?
Our idea is that SEDA Learning to Tutor Online participants (academic/academic related
colleagues) can create an AI generated visual metaphor image based on their
experience of being an online learner.
Participants choose one aspect of their learning (or an emotion - negative or positive)
that they have experienced on the course and generate an appropriate image (or other
visual media) to represent this in the form of a visual metaphor. Participants post this to a
shared discussion space, including a brief written explanation of their choice and a
proposed approach/solution that could be adopted for use with their own students.
This example contains Generative AI elements of:
•

Image

Why This Idea?
Having successfully utilised a visual metaphor task in the SEDA Learning to Tutor Online for
many years, in 2024 tutors adopted the use of AI image generation to enhance this multimodal, formative activity (CAST, 2024). The process of image creation (Watson & Barton,
2020), rather than sourcing an existing image, engenders deeper personal reflection
(Schön, 1983) and criticality in relation to their own experience of learning online and their
practice as an online tutor. Tutors encourage engagement from participants by creating
their own visual metaphor and adding to the discussion space (Bandura, 1977). The
insights generated from this activity supports participants to develop socially
constructivist, empathetic and compassionate pedagogies (Killingback, 2024) - the
foundation of our approach to online learning and teaching.

119


How Could Others Implement This Idea?
AI Visual Metaphor Activity
•

Create a shared discussion space/forum online (utilising your institutional VLE).
o

•

Example: Canvas VLE: Discussion

Clearly introduce the visual metaphor activity to participants, including links to
appropriate AI tools to support the task, e.g., Adobe Firefly: institutional licence and
Firefly has been trained on a dataset of licensed content, such as Adobe Stock and
public domain content where copyright has expired. Also include the rationale,
approximate time to complete the activity, due date and tutors’ involvement in the
activity.

•

Tutors model approach by creating their own visual metaphor utilising Adobe
Firefly, adding the image and accompanying text to the discussion space.
o

Example Gen AI image prompt: “An individual struggling to juggle too many
balls. They are dropping some of the balls but show determination to pick
up the balls and continue juggling.”

•

Tutors encourage participants’ engagement and provide formative feedback to
visual metaphor posts (with accompanying text). These actions foster a socially
constructivist online learning experience.

Links to Tools and Resources:
•

Canvas Discussions - Virtual Learning Environment: Online discussions/forums

•

Firefly - AI image generation tool

References:
Bandura, A. (1977). Social learning theory. Prentice-Hall.
CAST. (2024). Design Multiple Means of Action & Expression.
https://udlguidelines.cast.org/action-expression/
Killingback, C. (2024, July 9). To my students...please know that as your lecturer I care
about you: the importance of compassionate pedagogy. Advance HE.
https://www.advance-he.ac.uk/news-and-views/my-studentsplease-know-yourlecturer-i-care-about-you-importance-compassionate
Schön, DA (1983) The Reflective Practitioner: how professionals learn in action. Basic Books
Watson, B. and Barton, G. (2020). Using Arts-Based Methods and Reflection to Support
Postgraduate International Students' Wellbeing and Employability through
Challenging Times. Journal of International Students, 10(S2), 101-118.
https://doi.org/10.32674/jis.v10iS2.2849
120


Image:

Figure 1. AI Visual Metaphor Creation

Note. A university student generating a metaphor image using Adobe Firefly to share their
feelings of learning online (generated using Firefly).

121


Leveraging Generative AI for Enhanced Entrepreneurial Teaching
Name: Magdalena Plesa
Job Title and Affiliation: Lecturer in Social Entrepreneurship University of Liverpool
Contact Details: Magdalena.plesa@liverpool.ac.uk
Relevant Links:
Website - Magdalena Plesa

What is the MML Idea?
In this case study, I demonstrate an innovative teaching approach that integrates artificial
intelligence with traditional marketing persona templates. Specifically, I use AI to generate
visual depictions of target customer personas based on detailed textual inputs from a
standard persona template.
In practice, I ask students to complete a customer persona template and then use an AI
image-generation tool to create a visual representation of their target customer. This
multi-modality strategy not only reinforces the conceptual elements provided by the
template but also engages students’ visual and reflective learning modes. I also build in a
reflective component, where students present their personas and visuals to peers and
discuss the insights gained. The multimodal elements involved include text, image, and
speech (class discussion), allowing students to better understand and connect with the
personas they design.
This example includes Generative AI elements of:
•

Image

•

Text

•

Speech

Why This Idea?
This approach offers both theoretical and practical benefits by integrating AI-generated
visuals with traditional customer persona templates. Theoretically, it aligns with
experiential learning frameworks (Kolb & Kolb, 2005) and reflective practice (Gibbs, 1988),
allowing students to engage in a multi-sensory learning experience that bridges abstract
concepts with tangible representations.
Practically, it leverages generative AI to transform static information into dynamic visuals,
which enhances student comprehension and engagement. Preliminary evaluations
suggest this is effective, revealing that most students find the AI-generated imagery more
intuitive and engaging compared to text-only templates. For instance, student feedback
indicated that over half of the participants preferred the visual approach. This supports

122


the idea that the integration of AI fosters deeper understanding and a more immersive
learning environment (Chen, Ifenthaler, Yau, & Sun, 2024).
Previous research in entrepreneurship education has also shown that incorporating
innovative digital tools can improve learning outcomes and satisfaction (Béchard &
Grégoire, 2005). This evidence underscores the capacity of AI-enhanced methods to meet
the evolving needs of contemporary pedagogical practices. Moreover, this idea addresses
a clear and pressing need in teaching, learning, and assessment. Traditional teaching
tools often fall short in engaging modern learners, who benefit from dynamic, multimodal
educational experiences. Traditional teaching tools often fall short in engaging modern
learners. By combining structured persona templates with the creative power of GenAI, I
have found a scalable, adaptable way to enhance experiential learning across diverse
educational settings.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
1.

Introduce the Persona Template: Begin with a clear explanation of the customer
persona template and its importance in entrepreneurial strategy.

2.

Template Completion: Have students complete the persona template, focusing on
key demographics, goals, and identifiers.

3.

AI Integration: Guide students to use an AI image-generation tool (e.g., Dall-E) by
providing specific prompts derived from their templates.

4. Comparison and Reflection: Facilitate a class discussion to compare the traditional
template with the AI-generated visuals, encouraging students to reflect on both
methods.
5. Feedback and Adaptation: Collect student feedback and iterate on the approach
to refine its application in various teaching contexts.

Links to Tools and Resources:
•

OpenAI DALL-E

123


Image:

Figure 1. AI Generated visual of modern classroom

Figure 2. Customer personas template

124


References:
Béchard, J.-P., & Grégoire, D. (2005). Entrepreneurship education research revisited: The
case of higher education. Academy of Management Learning & Education, 4(1),
22–43.
Chen, L., Ifenthaler, D., Yau, J. Y.-K., & Sun, W. (2024). Artificial intelligence in
entrepreneurship education: A scoping review. Education + Training.
https://doi.org/10.1108/ET-05-2023-0169
Gibbs, G. (1988). Learning by doing: A guide to teaching and learning methods. Further
Education Unit.
Kolb, A. Y., & Kolb, D. A. (2005). Learning styles and learning spaces: Enhancing experiential
learning in higher education. Academy of Management Learning & Education,
4(2), 193–212.

125


Using NotebookLM Generated Podcasts to Develop Students’
Assessment and Feedback Literacy
Name: Samuel Saunders
Job Title and Affiliation: Educational Developer, University of Liverpool
Contact Details: samuel.saunders@liverpool.ac.uk
Relevant Links:
University of Liverpool - Samuel Saunders

What is the MML Idea?
NotebookLM is a Google product that is designed as a research organiser, assistant, notes
space and resource management system. It allows users to upload resources and
materials that are relevant to their research project and to use a GenAI-powered chat
space to interrogate, organise and interact with the uploaded content. However,
NotebookLM also contains (at time of writing) the option to create an ‘audio overview’ of
the uploaded resources, which constitutes an AI-presented podcast discussing the
content and synthesising conclusions about the material.
The audio overview feature of NotebookLM has the potential to be a useful source of
formative feedback for students completing assessments. A small scale pilot study was
undertaken on a Level 7 module on an English Literature programme, where students were
invited to upload their draft assessments to the software, generate an AI-presented
podcast, and listen to the output and reflect on what it says. Students were encouraged to
interact directly with the AI presenters, asking them questions and probing it for further
insights, with the ultimate goal of determining whether the assessments’ argument was
on the right lines, whether it had met the assessment criteria, and areas for development
ahead of final summative submission.
This example includes Generative AI elements of:
•

Text

•

Sound

•

Speech

Why This Idea?
Assessment literacy is of paramount importance to ensuring that students both complete
the assessment itself, but crucially that they actually learn something from the experience
(Smith et al, 2011). Smith et al argue that a meta-dialogue around (and about) the
assessment, its purpose and how it functions is therefore essential to help develop
students’ abilities to judge their own, and others, work, and thereby ensure that they
effectively meet the intended learning outcomes (2011). This activity is designed to help
126


develop this meta-dialogue, by providing an additional space where students can both
interrogate the progress of their own work by essentially interacting with it, and also
obtain a new perspective on what needs to be done to the draft submission to improve it
in the context of the defined marking criteria. Indeed, some students on this activity chose
to upload both their draft assessments and the marking criteria for the assessment, and
directly asked the AI presenters what needed to be done to the draft in order for it to reach
the higher levels of the mark scheme.
In addition, research has consistently shown that students perceive feedback more
positively when it is provided to them in an audio, as opposed to textual, form (Kirwan et
al, 2023). Ajjawi and Boud argue that feedback is inherently a communicative and social
act, charged with notions of power and emotion, and which therefore works more
effectively as a dialogue between multiple parties than it does as a one-sided perspective
where one party provides and the other simply imbibes (2018). It also helps to develop a
sense of belonging between students and staff (Killingback et al, 2019). This activity helps
to foster this sense of dialogue between students, but by adding a new party to the
conversation. Rather than it be a dialogue between tutors and students, or students and
each other, this activity is an example of students talking directly to their work, which, in a
manner of speaking, talks back.

How Could Others Implement This Idea?
1.

Provide access and instruction on the NotebookLM platform for students on the
module, and show an example podcast to demonstrate how it works if necessary.
Students must be familiar with the platform and sure that they are comfortable
using it. Outline the platform’s privacy settings to reassure students that it does not
violate copyright or IP restrictions, and ensure that there is an alternative process
for students who do not wish to use it - the activity should not be mandatory.

2.

Embed the activity on the module - outline a point where the students should
engage with the platform to the students themselves. It is an open question as to
what the best point is to complete this activity is on a given module, and it will be
dependent on local context, but a good rule of thumb is at a point where students
should have a substantial draft of their summative submission, but still have
enough time to make changes to the assessment based on the outcome of the
activity.

3.

Provide a space to reflect on the activity’s effectiveness. This can be a short Form
for students to outline the way they engaged with the podcast, or else students
could be required to simply submit their podcasts for you to listen to yourself, their
prompts or questions they used, or short reflective pieces on how effective or not
the activity was in helping refine formative assessments into summative
submissions.

127


Links to Tools and Resources:
•

Google NotebookLM | Note Taking & Research Assistant Powered by AI

References:
Ajjawi, R. and Boud, D. 2018. Examining the nature and effects of feedback dialogue.
Assessment & Evaluation in Higher Education. 43(7), pp. 1106-1119.
https://doi.org/10.1080/02602938.2018.1434128
Killingback, C., Ahmed, O., & Williams, J. (2019). 'It was all in your voice' - Tertiary student
perceptions of alternative feedback modes (audio, video, podcast, and
screencast): A qualitative literature review. Nurse Education Today, 72, 32–39.
https://doi.org/10.1016/j.nedt.2018.10.012
Kirwan, A., Raftery, S. & Gormley, C. (2023). Sounds good to me: A qualitative study to
explore the use of audio to potentiate the student feedback experience. Journal of
Professional Nursing, 47(1), 25-30, https://doi.org/10.1016/j.profnurs.2023.03.020.
Smith, C. D., Worsfold, K., Davies, L., Fisher, R., & McPhail, R. (2011). Assessment literacy and
student learning: the case for explicitly developing students ‘assessment literacy.’
Assessment & Evaluation in Higher Education, 38(1), 44–60.
https://doi.org/10.1080/02602938.2011.598636

128


Perk Up Your Academic Journey: A Coffee Shop Adventure
Name, Job Title and Affiliation:
Dr Laura Sharp (Senior Lecturer, School of Health and Wellbeing
Eric Davies (Learning Technology Specialist, Library Services)
Mia Wilson (Lecturer, School of Health and Wellbeing)
Ailsa Foley (Lecturer, School of Health and Wellbeing). University of Glasgow
Contact Details
laura.sharp@glasgow.ac.uk
eric.davies@glasgow.ac.uk
mia.wilson@glasgow.ac.uk
ailsa.foley@glasgow.ac.uk
Relevant Links:
University of Glasgow - Dr Laura Sharp
University of Glasgow - Eric Davies
University of Glasgow - Mia Wilson
University of Glasgow - Ailsa Foley

What is the MML Idea?
‘Choices at the Coffee Shop’ is an innovative, interactive, AI-powered tool designed to
educate students about academic integrity in a cost-effective way. We created realworld scenarios, which are presented in a gamified ‘choose your own adventure’ style
format, allowing students to understand the immediate and long-term outcomes of their
decisions in a consequence-free virtual environment.
This example includes Generative AI elements of:
•

Image

•

Sound

•

Speech

•

Text

•

Video

Why This Idea?
The game is embedded in an asynchronous teaching module: ‘Academic Values,
Originality, and Plagiarism’. Feedback indicates that 96% (n=50) agreed that ‘The module
has increased my understanding of plagiarism and how to avoid it’.
Cultural differences in perceived academic dishonesty can lead to unintentional
plagiarism (Fatemi & Saito, 2020) and unhelpful practices (Amsberry, 2009), requiring
anti-plagiarism pedagogies to accommodate diverse needs (Adhikari, 2018; Tran et al.,
2022). Gamification can accessibly enhance engagement through active learning (Khan

129


et al., 2017), positive feedback loops (Kaufmann, 2018), and a personalised learning
journey.

How Could Others Implement This Idea?
AI integration was a steep but valuable learning curve with unexpected and interesting
challenges. The design rationale highlights the approach’s adaptability. Feedback was
sought at each stage from students, ex-students, digital technology colleagues, and
teaching staff to optimise the usability and value of the tool.
Process

AI Use

Cost

Actions

Develop

None

N/A

Engaging and relevant scenarios were

Scenarios

created. AI was not used but may have
streamlined the process.

Relatable

None

N/A

Characters

Relatable characters were developed.
Jamie: from the UK, identifies as nonbinary
Naveed: cisgendered male with Indian
heritage

Character
Images

Dall-e-3
Photoshop

Paid

AI software was engaged to create

for

character images. Output required

Paid
for

considerable refinement. It became
apparent that generating diversity using
AI necessitated the challenging and
adoption of stereotypes. Refinements
addressed included character morphing,
removal of an earring, layering Naveed’s
clothing, and requesting Jamie’s tattoo
sleeve.

AI Voices

elevenlabs.

Paid

Perceived relatability and likeability of the

io

for

characters were considered when
creating the dialogue. The narrator was
initially overly formal, and one option for
Jamie’s became robotic.

Animating

Runway

Paid

AI developments allowed images to be

Images

Research

for

animated, however as Jamie lifted their

Gen-2

cup it elevated from the elbow rather
than the hand. Only a choice-point step
130


was animated with Jamie talking directly
to the camera.
Platform

Moodle
Rise

Paid

The game was embedded onto the Rise

for

platform within the learning module, and

Paid
for

this was then added as a SCORM
package to Moodle.

Links to Tools and Resources:
•

Advance HE Webinar: Offers an outline of the development process of the ‘Choices
at the Coffee Shop’ game, with illustrative examples of each of the steps above.

•

Academic Values, Originality, and Plagiarism: Full resource with ‘Choices at the
Coffee Shop’ game embedded in the ‘Academic Integrity and Values’ section.

References:
Adhikari, S. (2018). Beyond culture: helping international students avoid plagiarism. Journal
of International Students, 8(1), 375–388. https://doi.org/10.32674/jis.v8i1.170
Amsberry, D. (2009). Deconstructing plagiarism: International students and textual
borrowing practices. The Reference Librarian, 51(1), 31-44.
https://doi.org/10.1080/02763870903362183
Fatemi, G., & Saito, E. (2020). Unintentional plagiarism and academic integrity: The
challenges and needs of postgraduate international students in Australia. Journal
of Further and Higher Education, 44(10), 1305-1319.
https://doi.org/10.1080/0309877X.2019.1683521
Kaufmann, D.A. (2018). Reflection: Benefits of gamification in online higher education.
Journal of Instructional Research, 7, 125-132. 10.9743/JIR.2018.12
Khan, A., Egbue, O., Palkie, B., & Madden, J. (2017). Active learning: Engaging students to
maximize learning in an online course. Electronic Journal of e-Learning, 15(2),
pp107-115.
Tran, M.N., Hogg, L., & Marshall, S. (2022). Understanding postgraduate students’
perceptions of plagiarism: a case study of Vietnamese and local students in New Zealand.
International Journal for Educational Integrity, 18(1), 3. https://doi.org/10.1007/s40979-02100098-2

131


Improving Engagement with eLearning by Creating Multimodal
Video-Explainers
Name: Vivien Shaw
Job Title and Affiliation: Senior Lecturer, Northern College of Acupuncture
Contact Details: vivienshaw@nca.ac.uk
Relevant Links:
Northern College of Acupuncture – Teaching Faculty

What is the MML Idea?
I inherited an existing biomedicine course where students found the eLearning materials
to be long and difficult to engage with. In response to student feedback that they were
struggling to relate the eLearning to the LOs from in-person sessions and understand
what to revise, I used a combination of AI tools to create a transcript of each eLearning
component, then asked ChatGPT to create a synopsis including links to the learning
outcomes I had written for related in-person sessions. I then used the GenAI materials to
create a series of short ‘explainer’ videos that would allow them to navigate the lengthy
eLearning sessions with more ease. This was following the principles of Universal Design
for Learning (Fornauf & Erickson, 2020; Merry, 2024).
This example includes Generative AI elements of:
•

Image

•

Sound

•

Speech

•

Text

•

Video

Why This Idea?
The student feedback after our in-person sessions was that they enjoyed in-person
classes but were struggling to understand how the learning outcomes that I had
generated connected with the inherited eLearning teaching materials. This was
particularly important to them as their exam would be based on those learning outcomes
and a combination of their independent learning and what we had done in class.
I therefore took each eLearning video and created a transcript using the Transcribe
function in the Dictate area of Word on the web. This produced a transcript in Word, which
I then was able to put into ChatGPT to ask for a synopsis. I also entered the associated
learning outcomes into the prompt, and asked ChatGPT to create a structure for clearly
explaining the connections between the LOs and the content of the eLearning synopsis.

132


I then asked the AI to generate a slide deck from which I was able to create a 10 minute
‘explainer’ video summarising the eLearning, and relating it to the learning outcomes for
the session. The final recorded video included me talking around the topic, and was
illustrated using slides from the eLearning to help students to understand where the
information could be found and what elements would be particularly important as they
prepared for assessments.
This worked very well to reassure the students, and provide a solid structure for them to
study from. It did however also mean that some of the more strategic learners did not
engage with the longer eLearning resource as fully as they had before.
The principal outcome of creating these short ‘explainers’ from the long and difficult to
engage with video was that students stopped feeling overwhelmed and anxious, and
reported that they now felt that they knew what was expected of them.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
•

Provide clear, detailed steps to help readers implement the idea, including specific
GenAI prompts, the tools or platforms used, and whether they were free or paid
versions.

•

Make it adaptable for different contexts, ensuring it's simple and practical.

There were 4 different phases to going from a 1-hour eLearning class to a 10 minute
‘explainer’. They involved moving through multiple modalities, and eventually produced a
multimodal product including my narration and strongly edited GenAI slides/content.
1.

I used the transcribe function in Word ‘dictate’ on the toolbar to generate a
transcript of each eLearning session. This was free, but had a monthly limit of 300
minutes of processed audio. ChatGPT did volunteer various different options
including Otter.ai, Sonix, Descript. I looked at Otter.ai and hit a paywall, so used
Word because it was free. I did not investigate the others but think they could serve
a similar purpose.

2.

I then ran the transcription and the learning outcomes for associated in-person
sessions through ChatGPT to get a focused synopsis. An example prompt was:
“Create a synopsis of this eLearning material, and demonstrate where it fits with
the following learning outcomes: - Explain the different routes through which fluid
circulates from and to the heart; - Explain how fluid moves in a capillary bed; Apply your knowledge of vascular anatomy to acupuncture points ; - Discuss the
actions of core pharmaceutical drugs related to treating these conditions.”

3.

I asked ChatGPT to create a slide deck that summarises the eLearning material in
order to create an ‘explainer’ video of a maximum length of 10 minutes. The prompt
was “Use the transcript above to create a series of slides that will act as a
133


summary of the eLearning session.” There was material in the eLearning that was
relevant but not directly mapped onto the Learning Outcomes, so in my narrative I
used this summary to make those links apparent for students in a way that the AI
could not.
4. I created a 10 minute ‘explainer’ video based on the slide deck suggestions made
by ChatGPT, where I pulled out slides from the eLearning to support my narrative.
The video made regular reference to the slides in the eLearning where the topic
was discussed in more detail so that students would be able to navigate easily to
those sections of the eLearning.

References:
Fornauf, B.S., & Erickson, J.D. (2020). Toward an inclusive pedagogy through universal design
for learning in higher education: A review of the literature. Journal of Postsecondary
Education and Disability, 33(2), 183–199.
Merry, K., L. (2024). Delivering inclusive and impactful instruction: Universal design for
learning in higher education. CAST professional publishing.

134


The role of GenAI in Design for the Disassembly of Emergency
Shelters
Name: Farkhondeh Vahdati
Job Title and Affiliation: PhD candidate in Architecture, University of Liverpool
Contact Details: farkhondeh.vahdati@liverpool.ac.uk
Relevant Links:
LinkedIn - Farkhondeh Vahdati

What is the MML Idea?
In this project, I explored how multimodal Generative AI (GenAI) tools—particularly those
capable of generating visual step-by-step diagrams from annotated structural images—
can be applied in architectural education and emergency preparedness training. I used
GenAI to allow students to upload simplified, annotated diagrams of vernacular-inspired
shelters and receive AI-generated visual guides that illustrated disassembly sequences,
joinery systems, and the hierarchy of materials and components.
By applying this process, I supported students' spatial understanding and engagement
with concepts such as design for disassembly (DfD), modularity, and sustainable
construction practices (Vahdati, Tedjosaputro, & Damavandi, 2025). The AI-generated
visuals helped translate complex 3D structures into sequenced 2D guides, reinforcing
spatial literacy and multimodal communication in architecture education.
This example includes Generative AI elements of:
•

Image

•

Text

Why this Idea?
I chose this approach because the use of GenAI in architectural education aligns with
constructivist learning theories, where students co-create knowledge through iterative
feedback and visual exploration. Tools like GPT-4 helped students better comprehend and
communicate spatial relationships—skills that are often difficult to grasp through text
alone. Research by Bower et al. (2022) has supported this, highlighting the ability of AIassisted tools to enhance cognitive engagement in design education, where multimodal
understanding is key.
Beyond higher education, I also considered a broader application: helping non-specialists
understand how to assemble or disassemble emergency shelters during crises. Using a
custom-trained AI model informed by emergency design standards (such as UNHCR
guidelines), I explored how AI could produce accessible, language-free visual guides to
support disaster response and recovery. This bridges humanitarian design with academic
135


training and makes complex architectural knowledge transferable to real-world settings
(Vahdati, Tedjosaputro, & Agkathidis, 2025).
Ultimately, the multimodal nature of this approach (image + text + interaction) responded
to a growing demand for adaptive, visual learning in education (Jinuntuya & Theppipit,
2007), while also demonstrating how AI can shift from automating outputs to co-creating
and enriching learning experiences.

How Could Others Implement This Idea?
Step-by-step instructions for implementing the idea.
1.

Select source material:
Start with annotated diagrams or clear, labelled images of simple structural
systems—such as vernacular shelters—that highlight joints, sequences, or material
layers.

2.

Use a multimodal GenAI tool:
Upload these diagrams to a platform like ChatGPT with GPT-4 & Vision (ChatGPT
Plus). Use prompts like:
“Generate a step-by-step visual disassembly guide showing material layers, joints,
and sequence, based on design for disassembly principles.”

3.

Refine and compile outputs:
Use Canva, PowerPoint, or similar tools (free versions are sufficient) to organise
and annotate the AI-generated diagrams into visual guides or slideshows.

4. Integrate into teaching:
These resources can be used in architectural studios, humanitarian design
modules, or emergency preparedness workshops. Depending on the audience, the
material can be simplified for lay use or expanded for advanced student tasks
such as critique, redesign, or reconstruction exercises.

Links to Tools and Resources:
GenAI Tools (Visual + Text-Based)
•

ChatGPT with GPT-4 & Vision (ChatGPT Plus)
Use to analyse images and generate step-by-step descriptions.

•

Canva (for compiling visual guides)
Use free version to create instructional diagrams or slide decks.

136


Image:

Figure 1. Screenshot of Prompt for a Step-by-step visual disassembly on ChatGPT

137


Figure 2. Step-by-step visual disassembly by ChatGPT

Figure 3. Design for the disassembly of a shelter using ChatGPT4o

138


References:
Bower, M., Nguyen, G.N.H., & Stevenson, M. (2022) The discourse of design: Patterns
of TPACK contribution during pre-service teacher learning design
conversations. Education and Information Technologies, 27, pp.8235–8264.
Available at: https://doi.org/10.1007/s10639-022-10932-w
Jinuntuya, P., & Theppipit, J. (2007). Temporary housing design and planning
software for disaster relief decision support system. Proceedings of the 12th
International CAADRIA, 639-644.
Vahdati, F., Tedjosaputro, M., & Agkathidis, A. (2025). AR-Assisted 3D Generator: A
Parametric Investigation into a Custom Emergency Shelter Design. eCAADe
2025-Confluence
Vahdati, F., Tedjosaputro, M., & Damavandi, F. (2025). Comparison of AI-Driven and
AR-Driven Design for Disassembly (DfD) of a Shekili-inspired Emergency
Shelter Design. The 5th International Conference on Artificial Intelligence in
Education 2025 (ICAIE 2025)

© 2025 by the University of Liverpool, Centre for Innovation in Education.
An Educators’ Guide to Multimodal Learning and Generative AI is made available under a Creative
Commons Attribution Non Commercial 4.0 International License.

139