---
source_url: https://arxiv.org/abs/2510.01467
ingested: 2026-05-07
updated: 2026-07-28
version: v2 (last revised 2026-07-24)
sha256: e79233e7991c5bd4ec2d61ba9ead056c6b00196387970c997c6e5f94b647ee1a
---

A principled way to think about AI in education:
guidance for educators and policy makers on
action based on goals, models of human learning, and use of technologies
Noah Finkelstein
University of Colorado, Boulder
Drafted 7/30/2025
with an Update 7/24/2026
The rapid emergence of generative artificial intelligence (AI) and related technologies has the
potential to dramatically influence higher education, raising questions about the roles of
institutions, educators, and students in a technology-rich future. While existing discourse often
emphasizes either the promise and peril of AI or its immediate implementation, this paper
advances a third path: a principled framework for guiding the use of AI in teaching and learning.
Drawing on decades of scholarship in the learning sciences and uses of technology in education,
I articulate a set of principles that connect broad educational goals to actionable practices. These
principles clarify the respective roles of educators, learners, and technologies in shaping
curricula, designing instruction, assessing learning, and cultivating community. The piece
illustrates how a principled approach enables higher education to harness new tools while
preserving its fundamental mission: advancing meaningful learning, supporting democratic
societies, and preparing students for dynamic futures. Ultimately, this framework seeks to ensure
that AI augments rather than displaces human capacities, aligning technology use with enduring
educational values and goals. It is meant as a practical guide for anyone engaging in the use of
new technology tools in their educational practices, or those people setting educational policies
in the modern era.
A new preamble (Jul 2026) and modest updates throughout the paper contextualize it since
initial sharing (Jul 2025). The good news is that the arguments and principles of action posted
originally appear to still be relevant one year later.

Preamble:
A, or perhaps the, critical question that faces us with the regard to the use of generative artificial
intelligence and related technologies (AI) is:
In an era when computers and robots can engage in (or appear to humans to
engage in) all human activity as effectively or more effectively than humans,
what are those skills, habits of mind, and practices that we must maintain as
within the human purview, in order to maintain our humanity and societies?
The question begins with an imagined future, and for the purposes of this paper and its
associated recommended practices in education, we may not need to determine whether this
possible future does or could exist.1 No matter the answer, this future is worth considering, both
because it is within the realm of possible futures and because such a future ought to inform our
current practices. The question concludes by framing the goals -- or what is at stake currently -for our humanity and society.2 The concluding scenarios of this piece examine some short-term
outcomes (e.g., a much more highly automated educational system); these are only
steppingstones to the longer-term cultures we will build. And, finally, central to the question is
what we as educators and policy makers have control over: defining what we do educationally the skills, practices and habits of mind, directed toward the possible future we seek.
While the question is essential for our society as a whole to consider, we in education have a
unique position to play in answering. As just noted, we have the opportunity to define and
develop the practices, skills, and habits of mind of individuals in our society. In fact, this is why
we constructed our educational system, and why it is so contested politically and in public
discourse, today. We have the privilege and responsibility to act for the welfare of individuals
and the public good, as will be discussed in more detail below.
While the role of the educator and the learner is examined in more detail below, three objectives
for our educational practices stand out as essential. We need to focus on cultivating:
discernment: the abilities to ask and frame appropriate questions, validate and contextualize
answers to these questions, and to act appropriately with the results.
empathy: the ability to understand and share the perspectives of others, which underpins our
abilities to engage in communication and collaborative work
sense of self: where students learn about their roles, purposes and whether and how they
belong within a course, field, or broader cultural system.
These reiterate some of the thinking below, but are particularly important in the modern era.
Accessing information, results of calculations, and analyses have become nearly instantaneous
with modern technologies. Discernment develops not only abilities to engage in these practices,
but essentially to ascertain the legitimacy, utility, applicability of the outcomes we get. It also
informs what questions we ask, those that give rise to this information seeking behavior. Lastly,
1

Whether AI is simulacrum or engages in genuine forms of human activity, such as thought and empathy is an
important question and long debated (e.g. Searle, 1980). Certainly, computers are different from people. However,
the tasks, interactions, and forms of behavior they engage in can pass as human, and surpass human practices in
many areas already. Their capacities, at the time of writing, continue to accelerate.
2
Here, I am not seeking to be hyperbolic and consider the world turning into paper-clip factories (Bostrom 2003),
but rather to argue that we have collective agency in defining what it means to be human and what societies are.
Similarly, Pope Leo XIV has recently issued an encyclical examining these same matters (Pope Leo XIV, 2026).
Finkelstein, A principled way to think about AI in education

1

discernment shapes how we act based on the outcomes. Collectively, the practices of
discernment support the development of metacognition (reflecting and regulating one's own
thinking), and ethics, morals and judgement, as appropriate to the disciplines, communities, and
society. Another key outcome and practice of education is our abilities to engage in collective
work, and productive engagement in a cultural system (whether discipline or broader society);
empathy underpins these capacities and thus, itself, is an essential practice and outcome for our
educational system. Balancing our collective capacities is the development of self and sense of
belonging. In our systems, learners can develop agency and purpose. These qualities underpin
broader academic success, including performance, engagement, and retention within fields.
Both the question and these answers are explored in more depth throughout this essay, providing
key tactical choices we get to make as educators and policy makers. Here, I focus on the what choices we individually make for our educational practices. As we seek to understand how to
engage at scale, we must draw from models of institutional and organizational change in higher
education – an inquiry beyond the scope of this piece. Notably, we have made enormous
progress on this topic in recent decades (e.g., Henderson, et. al, 2011 and NAS, 2024). However,
current models of institutional change generally presuppose our goals focus on adoption of
proven, evidence-based approaches to educaiton. In the case of AI-infused educational practices,
our evidence-base is much sparser, and yet they are arriving in our spaces anyhow. Hence, our
models of educational transformation need to be updated. In a recent piece, we provide a
framework and design implications for change initiatives to consider in the modern era (PerlNussbaum, 2026).
Introduction:
Why a Principled Approach to AI in Higher Education Teaching and Learning
If generative AI and associated new technological tools can deliver
instruction to our undergraduates in a just-in-time and studentspecific manner (Kestin, et al 2025), why would our institutions of
higher education continue (given their financial models and
founding missions are based on undergraduate education)?
If students can use generative AI to produce essays, homework
solutions, and project designs without effort (Lohr, 2025), why
would they spend the time learning (instead of more cheaply
producing these products that are used to evaluate them)?
If these new tools change how we think and conduct work in our
academic disciplines (Kosmyna, 2025), how do we decide what we
should do and what forms of thought are valuable?
As generative artificial intelligence (AI) and other technological innovations are becoming more
widespread, are rapidly evolving, and likely to substantially change our educational system, it is
worth developing key principles about the use of these tools to support learning and our learning
environments. While a myriad of work has focused on (1) the promise and perils of these
Finkelstein, A principled way to think about AI in education

2

technologies, (Capraro, 2024) and (2) ways to implement these technologies in our learning
environments (Mollick, 2025a, 2025b) here I seek to promote a third way - a foundational
approach to the above – principles to help us think about the productive uses of new technologies
in education. These principles build on decades of scholarship of teaching and learning and of
technology use, and can inform how we use of these new powerful technologies to support our
educational goals.
While considering both the largest scale vision (promise and peril) and on-the-ground practical
steps (implementation) of technologies in education are necessary, a set of guiding principles can
bridge the gap, connecting our implementation approaches to the large-scale goals we have for
education. Many high-quality implementations and studies of AI in education are helping shape
our use of these tools in the hopes of achieving broader goals. Each of these applications have
principles that underpin their use; however, such principles are all-too-often implicit. That is, for
example, when providing details on the use of generative AI as a tutor for students learning in
college physics, or training students on prompt engineering for supporting the development of an
argumentative essay, a variety of guiding principles underpin these pedagogical approaches, but
often remain unstated (e.g., how are the tools used to support interactive engagement, balance
cognitive work between human and computer, etc.).
Not only do principles of use allow us bridge between high level goals and on the ground action,
but they also give us tools to ethically adapt, experiment, and implement as these tools evolve.
We are at the beginning of a grand disruption in our fields; committing to principled-based
choices and actions that are based on decades of scholarship will allow us to bootstrap, invent,
learn, and revise our approaches while maintaining the promises of a relevant, moral, and
effective educational system. By more thoroughly articulating guiding principles, we can: seat
technology-based educational practices in the broader constellation of educational activities;
ensure that these new approaches align with, and in fact accelerate our capacities to realize,
rather than undermine our goals; create adaptable, ethical approaches that evolve with the
technologies; and develop a robust strategy for the sustained and successful implementation of
AI and other technologies.
These technologies are particularly rapid in its evolution and deployment. And, much like AI
itself, where AI is headed is hard to predict. To that end, while it is hard to make predictions,
especially about the future (Steincke, 1948), we might do well to skate to where the puck is
going to be rather than where it is or has been (Gretzky, W., n.d.). We have grand agency and
capacities for shaping the future of education, And, if we do not proactively and engage in the
use of these new technologies for education, they will still be present. They will still be used.
However, they are more likely to use us - to shape our interactions with each other and our
students that we did not intend, than if we had sought to intentionally engage. With appropriate
forethought, these new tools can help us advance our capacities and the capacities of learners to
realize the future as it ought to be.
In seeking to productively develop and engage with these new tools in our educational
environments, it is worth considering some basic questions:
-

What are our goals in education?

Finkelstein, A principled way to think about AI in education

3

-

Given practical and actionable definitions of education and of technology, how do we
consider appropriate roles for educators, learners, and technologies?
How will the use of these technologies transform what we are learning?

This essay seeks to explore these questions to provide the reader a set of principles by which
they may bridge between implementation strategies and realize the promises and avoid the perils
of these new technologies in education. These principles offer guidance for those seeking to
understand and deploy new technologies in their educational environments, both formal and
informal. Because our educational contexts vary dramatically (and in fact are a strength and
hallmark of the American education system), these principles are designed to be flexible and
help allow us to contextualize the use of these technologies to the specific environments
educators find themselves in, as well as to allow for a variety of goals of education. The present
focus is on colleges and universities; though, many of the approaches will likely apply to preand post-college environments.
Why educate?
While there are many excellent volumes written about the purpose of education (e.g. Dewey,
1916), starting with such a fundamental question is essential in ensuring that our educational
practices are aligned with our goals. Three broad classes of goals span the space of why we have
constructed our educational systems:
1) Developing individuals. Learning is the mechanism by which people develop (Vygotsky
1978). The development of higher order cognitive functions not only supports basic skills
for engaging in society (language and mathematical literacies, reasoning, argumentation,
communication, socialization, empathy, agency, identities, and more), but also makes
individual’s life more meaningful, richer, healthier, and connected.
2) Societal infrastructure. The roots of most modern societies are grounded in an educated
and informed populace. Democracies are dependent upon basic skills and common
culture developed through education. As our societies become more interconnected (both
within and across communities), as the decisions that one collective makes increasingly
impact other collectives, education provides capacities to understand and address these
interconnections in ways that are to the benefit of all.
3) Workforce development. The most common, modern framing of the purpose of education
is to get a job, or more appropriately, to support a career. Arguments around workforce
development vary from specific skills training and certification to gaining broader skill
sets and supporting the capacities for job and place-based learning. Motivations also vary
from fulfilling the needs of industry and the economy to fulfilling the interests and needs
of our learners.
Principle 0: Know why you are acting -- whether at the highest level or specific class activity.
What are the goals you are seeking to advance? Ultimately the goals one prioritizes ought to
inform the actions one takes and how one may productively leverage new technologies. These
goals can be multifaceted, and potentially contradictory, as happens regularly in our education
systems.
The list of goals above is placed in (my) priority order. While workforce development and job
training goals are the most commonly cited, they are achieved by addressing the initial two goals
- developing learners and preparing them to engage in society (AAC&U, 2009; Finley, 2021).
Finkelstein, A principled way to think about AI in education

4

Developing basic reasoning and communication skills and understanding historical and
humanistic traditions provide students the skills useful for today’s and tomorrow’s jobs.
However, preparing students for specific workforce needs and associated skills, does not
necessarily prepare them as active and engaged citizens, nor to develop their own perspectives
and passions. Others may have other goals for our educational system. No matter the objectives,
being explicit about our goals ought to guide the use of technologies in creating our educational
environments. Will technologies support the development of more self-realized, empathetic,
community engaged and connected, rational, civic, and artistic individuals?

What is learning? and teaching?
Again, volumes have been written about what learning is and how and where it happens. And
while there are perfectly valid definitions of learning based on biology and brain changes, e.g.
learning is the creation of neural pathways, such definitions are less operational than
complementary perspectives drawing from socio-cultural theories of learning. Learning is about
socializing humans into cultural systems (Vygotsky1978 / Cole 1996).
Learning is the internalization of culturally developed knowledge and practices.
It occurs through the interaction of humans with other humans directly and
indirectly, through the use of tools.
Thus, culture shapes learning and vice versa. Whether a child learning arithmetic - internalizing
what 6x 4 = 24 means (e.g. six groupings of four objects) along with the associated algorithms,
or what the four freedoms enshrined in the First Amendment of the U.S. Constitution are and
how they apply today, or how computer-aided-design software may be use in the manufacture of
assistive technologies, people learn material deemed important by cultures through interactions
with other more knowledgeable peers, culturally relevant tools, and their proxies.
An appropriate parallel consideration is where learning ought to occur – i.e. in formal settings
(schools / colleges), on the job training, or informal and community settings. However, such
considerations are beyond the scope of this piece, which will focus on institutions of higher
education and assume their value and relevance (for the moment).
What are these advanced technologies?
When thinking about technology it is easy to immediately think of new “hi-tech” tools and
silicon-based machines. And, while these indeed are technologies, it is productive to broaden the
lens so that we consider other technologies and humans’ long history of developing and using
these technologies -- for example, pre-silicon-based technologies include: a hoe, a pencil,
language, and or our disciplinary domains, such as physics. Notably these technologies (or tools)
are both material (a hoe) and intellectual (physics). A functional working definition is useful in
considering technologies in education (Cole, 1996; Farrell, 2025a, 2025b):
Technologies are human constructed tools, both material and intellectual, that reorganize
how humans interact with each other, with the world, and with other technologies.
Consider a pencil and paper – these technologies modify both human memory (expanding our
capacities to track a larger number of data bits than our short-term memory tracks alone) and our
Finkelstein, A principled way to think about AI in education

5

abilities to communicate (expanding communication across space and time). Of course, Socrates
has a point that writing may come at the detriment of our internal memory systems and possibly
oral communications. Similarly, language systems or disciplinary fields change how humans
interact. And of course, generative AI, LLMs, machine learning, and other rapidly evolving
computer-based tools reorganize human practices and capacities.3
Deploying these new technologies for education.
Drawing on these actionable definitions of learning and technology, we might consider how
humans use these new tools to productively support our goals of education. A natural response is
to consider: what are these new technologies good at compared to what are humans good at, and
subsequently allocate responsibilities based on these capacities. For example, until the early
2000s computers, which had been around for decades, were very good at computation and poor
at vision (parsing images). As such, much of computer work had replaced manual computation
by humans and humans remained in charge of discernment in image production and detection.
In recent years that has changed, and notably computers have become very capable of image
detection and processing - often and initially with significant errors. But this scenario raises the
questions:
Just because a technology is capable of a task, should it be assigned to and responsible
for that task?
If new systems can teach, should we replace educators?
If these technologies can learn, do they replace students?
Of course, these are hyperbolic and provocative framings. Rather than such an absolutist and
reductionist framing, a potentially more productive approach will be to consider how these new
and emerging technologies might advance human capacities.
Principle 1: In our educational enterprise consider what domains of work are necessary or
essential for educators to maintain as human led and directed, and what is feasible and
safe to outsource to these new technologies.
Principle 2: Determine what practices and activities are necessary or essential for
students to engage in as human led and directed, and how newly advancing technologies
might advance learners' capacities.
Notably, these principles do not prescribe binary actions. There are degrees of responsibility to
be allocated, and simultaneously, new responsibilities will emerge, because of the development
of technologies. To that end, we ought to consider what are the current and near-term roles and
responsibilities for educators and learners.
Roles of the educator

3
Technologies are typically neither good nor bad- it depends upon how they are used. Though, notably,
technologies do come with predispositions for use, which embed values. Pope Leo XIV has an excellent
examination of this point in his encyclical on AI (Pope Leo XIV, 2026).

Finkelstein, A principled way to think about AI in education

6

There are a myriad of roles that educators serve in advancing learning, and, yet again, volumes
have been written about the roles of teachers (NAS 2025). Here, zooming in one layer further
from the goals of education that we opened with (individual development, societal infrastructure,
and workforce needs), we might consider what are the classes of activities that an educator (in
higher education) engages in and has control over. Here, we consider them from a design or
reverse engineering perspective. Knowing what our goals are, how do we achieve these ends?
And, how do we know if we have achieved them?
The following are key classes of action that an educator engages in – they ought to shape and
will be shaped by the uses of technology. These roles include:
Establishing the objectives of a course, units, and lessons. While not established in isolation, it
is the instructor who contributes to and enacts activities with various goals, both explicit and
implicit, in our classrooms.
Content, Practices and Skills development. These are the dominant goals, or learning
objectives, in our classes. Does a student know what Newton’s Laws are, and when and
how to apply them by developing algorithmic proficiency? This may also include
defining the bounds of a domain, e.g., what are and are not considered “physics
questions”.
Ways of thinking, habits of mind, and metacognition. These are the tools that we use in
developing and applying our content understanding and skills. We educate students in
the habits of mind of a field, the appropriate intellectual moves, and methodologies.
Simultaneously, we teach students to think about and regulate their own thinking
– metacognition. Finally, these habits of mind may include discernment, ethics, and
morals within a field or cultural system, or society more broadly.
Motivation, empathy, trust, identity, purpose, and community. The educators and
community of learners in a course contribute to individuals’ commitments - personal and
collective engagement in the field. Within a course, educators cultivate participation (or
not), build a sense of trust, belonging, and identity of individuals within the field (or not),
and make the field relevant for individuals and broader communities, frame the purpose
of learning and the domain, and build a sense of community within the class and domain.
Many of these characteristics are underpinned by and developed through empathy, which
can be cultivated (or curtailed) in our classes.
Principle 1.1: identify which goals your classes have, and make these goals explicit in
educational practices, through the use of modern technologies (including generative AI).
Note which goals may be subverted through the use of generative AI.
For example, is learning syntax of a programming language essential or outmoded? And
how might the use of new technologies support or circumvent goals for deeper reflective
thought in computational problem solving, or cultivation of more members of our
communities engaging in computational thinking? Is writing an essay "independently"
necessary? And if not, how do we ensure the students develop attention and emphasize
clarity of thought?

Finkelstein, A principled way to think about AI in education

7

Curating information, knowledge, and resources of a domain. An instructor curates learners’
access to information, knowledge and disciplinary tools and practices that are in service to the
goals of a course.
Information. Historically, education occurred in an information-limited age. An educator
provided access to information. A Personal library (having access to information) was the
sign of an educated individual. In relatively recent decades, we moved to (more or less)
ubiquitous access to information. The role of educators has, in fact, now been shifted to
limit, vet, and validate the information students attend to, to assist them in focusing their
attention on relevant information.
Knowledge. The organization of information, or knowledge, is one of the desired
outcomes of education, and a form of content in a domain. An educator curates
knowledge, organizing the information that students have access to, so that these
knowledge structures allow a student to productively integrate and assimilate additional
information and develop insights.
Resources. An educator provides access to essential material and intellectual tools and
governing practices around these resources. Whether laboratory equipment or library
search tools, a calculator, or perhaps generative AI engines, an instructor introduces,
curates and guides learner’s use of domain-appropriate tools and practices, and their
application.
Principle 1.2: Consider how the curation of information, knowledge and resources may
be augmented and needs to be adapted in light of new technologies, such as generative AI
tools.
For example, and perhaps most obvious, how should generative AI be used within a
domain and how should students be taught to use it? If new technologies begin curating
information, and indeed knowledge structures, how can instructors utilize these resources
productively? I would love for generative AI to replace the clicking I currently do in
creating and updating my Orwellian-named learning management system.
Commensurately, how should students be taught to use (and validate) technology-based
information, knowledge, and resources, not only for this course but as a foundational lifeskill?
Design of the educational activities (and possibly the environment) appropriate to the learners
who arrive. A great deal of scholarly work in recent years has attended to pedagogical
approaches and curricula for learning -- both within and outside our classroom environments.
Key principles in the design of educational activities are nicely summarized in a recent
consensus study from the National Academies (NAS, 2025).
Actively engage students. Decades of evidence and foundational theories of learning note
that learning is an active process (not the absorption of information), and environments that
proactively engage students show demonstratively higher learning gains than our
traditional, one-way, lecture environments.
Leverage students’ backgrounds to motivate and engage. Students’ varied interests, goals,
knowledge, and experiences impact what and how students learn. Support the affective,
social, identity, and community development of learners. Whether a student engages or not,

Finkelstein, A principled way to think about AI in education

8

whether they learn or not, will depend upon how they feel about the domain, their
identities, and sense of belonging.
Build adaptable and transparent environments. Sharing the goals, processes and outcomes
with learners and the communities will ensure more students are more successful. Adapting
to student interests and needs is essential if education is to be seen as relevant and
necessary.
Principle 1.3: Consider how technologies may assist in each of these design components of
effective instruction. What are the benefits and liabilities of replacing or augmenting
instructor activities with AI? How might technologies assist in the selection of effective,
educational activities that are responsive to the students in a particular course?
For example, many basic, skills-based activities may be served by a computer tutor (think
typing-tutor from a certain era), and while it might be tempting to personalize all
interactions for a given student (Stephenson,1995), such activities may undermine a goal of
socializing a student and helping them regulate their own learning, abilities to interact with
others, and navigate uncertain and complex situations. In a complementary approach, it
may also be the case that AI can serve as a concierge of sorts for faculty in designing their
educational environments.
Assessment and Certification. A core purpose of education is assessment and certification of
individuals' educational outcomes and evaluation of the educational approaches themselves.
Assessment of past, current and future states of learners. To be effective, educational
activities need to meet students where they are, building on what they already know -hence, assessing the prior state of student learning is key. Historically, this has occurred
through course sequencing and prerequisites. More modern (and expensive) approaches
include assessment of students on entry to a program or class. Assessment of the current
state of the learner can be developmental (formative assessment) both for the student and of
educational practices themselves. Often, this formative assessment occurs through lowerstakes assignments, such as homeworks and quizzes. Assessment of the end-state of
learners in a class, most often represented by student performance on exams, essays and
other activities, and ultimately a course grade, is the summative form of evaluating
students. In some instances, and potentially very valuable, we can measure the future
trajectories of students, documenting what they will know and be able to do in the future
(Vygotsky).
Certification. Ultimately instructors certify the understanding of students enrolled in their
courses. They certify that a student is ready to continue, or not, in a course sequence.
Students, future educators, potential employers and others in society have vested interest in
certification of individuals, especially as this certification captures student understanding,
skills, practices, and the goals presented above.
Principle 1.4: Consider in what ways technologies may assist in the assessment and
certification of individuals. In what ways do humans need to exercise professional
judgment and to what degree might this be outsourced to a machine?

Finkelstein, A principled way to think about AI in education

9

For example, our current summative assessment and certification is quite coarse,
symbolized in a grade, certificate, or degree. It represents aggregate assessment of skills,
knowledge, ways of thinking, and proficiencies. That is, grades are not invertible
functions.4 Technologies may be able to provide a compact accessible representation of
student capabilities in given circumstances that may be unpacked, or delved into, as it suits
the interests of the learner and others seeking to work with the learner. Many novel efforts
are seeking to expand and tailor our capacities as educators to provide individualized
actionable feedback, at scale for our classes. These features may become particularly
important roles as we develop more certificates, badges and micro-credentials.
Assessment practices and continuous improvement: Based on the educational outcomes of
learners, and other considerations (such as resources required and institutional constraints),
an educator ought to engage in a process of continuous improvement informed by the
specific context, student outcomes (and student input), and by scholarship in the field (NAS
2025).
Principle 1.5: Consider how to leverage technologies to inform curricular and pedagogical
adaptations based on student assessment.
For example, an instructor might leverage generative AI to curate scholarly and evidencebased recommendations for educational practices to address a given situation. And, there
may be an opportunity for technologies to analyze the breadth of student performance data
in a course to provide summaries and potentially, even recommendations for tailored
pedagogical approaches.
Supporting Learners' Roles
Students' roles may parallel and complement those of the educator, listed above. Glibly, if the
role of an educator is to teach, the role of a student is to learn. If an instructor dominantly defines
the goals of learning, students can simultaneously contribute to these goals and contextualize
them for themselves. If an educator is to curate information, knowledge and tools, the student
ought to access, use, and internalize these. If an educator is to actively engage students, the
students are to engage. And so forth. The following takes a learner-centered perspective and can
help inform actions for the student, instructor, and educational leaders.
Attending to goals: What are the necessary approaches a student ought to engage in to attend to
the highest-level goals listed in the introduction? What are the student goals for participation in
an activity, course, or degree program? If a student is seeking a job or certification, their focus
may be on content, and associated skills, for example, knowing how to engage in computer aided
design, coding or developing technical writing skills. If a student is seeking a career (multiple
jobs) and to join a community, their focus may additionally include understanding ways of
thinking, discourse patterns, norms, and habits of mind of the community. If a student is seeking
to define a community, their focus may additionally include attention to developing
metacognitive skills, discernment, ethics, morals and purpose. And personal development of an
individual may additionally involve the attention to identity development, motivation and
purpose.
4

Many thanks to Edmond Johnsen, CU Boulder Center for STEM Learning, for this framing.

Finkelstein, A principled way to think about AI in education

10

Principle 2.1: Ascertain, attend to, and develop the goals our learners. What are
appropriate and needed goals in our future, given the new capacities and trajectory of
technologies?
For example: if a student arrives with the intention of only learning specific skills, we can
provide opportunities to develop those skills, potentially utilizing new and emerging
technologies, such as computer-tutors. At the same time, we might help make students
aware of broader opportunities of education and engage them in understanding the future
roles of technologies in their lives.
Practicing competencies and acting on assessments: Each of these goals will have related
practices, skills, and competencies that a student may develop. A student will need to engage in
routine practice, with feedback to develop the associated content and skills. How does a student
respond to the feedback provided (whether formative and summative)?
Principle 2.2: Carefully consider which practices students are engaged in (both those they
are asked to engage in and those practices beyond what they are asked to engage in), and
ascertain in which ways generative AI and new technologies may assist and may
circumvent the goals that the students and the instructor hold.
For example: A great deal of recent attention has been paid to students' use of generative
AI in computer coding classes (Lohr, 2025). In many courses, some students are producing
code through the use of AI tools, rather than producing code themselves, as required by the
course. Here, some of the challenge is that student performance is not aligned with learning,
and may not be aligned with instructor goals. Similar challenges show up in the humanities
(O'Rourke, 2025). Can we support students' clarity of thought and expression, develop their
values and voice, and minimize the temptation to outsource thinking and attention? How
might we adapt our educational activities to address these challenges? Do we regulate
students (holding in-class, supervised assessments), or do we modify the goals and
activities that students engage in to prepare them for a new technology-laden landscape?
Applying and Synthesizing: In our current system, much of our educational practice relies on
the student to synthesize across courses, to contextualize and apply this understanding to their
own lives and trajectories. Of course, there are some courses that look to span disciplinary
content, e.g., capstone classes and internships provide some structured opportunities for learners;
however, these are far from the norm.
Principle 2.3 Consider the ways new technologies might assist or circumvent the
opportunities to analyze and synthesize across courses and disciplines.
Example: There is some movement to have student generated portfolios, using new
technologies. Such portfolios can span across courses and serve to complement student
grades and instructor and institutional certification, demonstrating what students can do.
Of course, the role of authorship – to what degree is a portfolio student versus AI
generated – will be a matter for all to consider.
Finkelstein, A principled way to think about AI in education

11

Collaboration, including co-design of educational environments. Whether implicitly or
explicitly, our educational environments teach students how to collaborate and contribute
collectively (or not). And, students play an essential role in the design and enactment of our
classes themselves. Most often students’ roles in the co-design of educational experiences are
implicit. That is, students can vote with their feet, participating in various elements of a course,
and in practice, to grow or shrink demand. Increasingly, while attending to goals, educators are
spending some course time working with students to explicitly design the educational
experience, to ensure that it is relevant, meaningful, and impactful. Whether establishing
collective course norms, for example around cheating or use of AI, or around content, allowing
students to define, to some degree, what is covered in a course and how it is covered, students
can partner with and make more effective learning environments.
Principle: 2.4 Review the roles that students might serve in collaborative design and
enactment of our educational environments. Instructors and others can work with students
to consider how generative AI might advance the goals and support the practices (such as
those established in 2.2) in our classes.
For example, we can consider working with students to define what roles generative AI
and other novel technologies might serve in a class. Can students co-generate a set of
policies and governance practices in class round the use of AI as part of class activities?
Additionally, an instructor might work with students to identify educational practices that
prevent the nefarious uses of generative AI, e.g. those that circumvent the learning goals,
and instead proactively align performance measures with learning and understanding.
Review our educators / system. Students, dominantly through rating systems, evaluate the
instructor and course for institutional and individual feedback. These reviews occur through
formal college-sponsored mechanisms, third party sites, or word of mouth. They also implicitly
rate a system based on their participation - do they enroll? As our institutions become more
dependent upon student tuition as a funding source, such explicit and implicit ratings will only
increase in value. Notably, however, our current rating systems suffer from not being coupled
(correlated) with the instructor-defined goals of our course, nor to measures of faculty effective
practices, nor to the highest-level objectives and goals listed at the beginning of this piece.
Principle 2.5: Consider the roles that new technologies might serve in supporting
students’ evaluations, as well he collection of evidence and their analyses used for
evaluation.
For example, building on longstanding adaptive technology use, these technologies can
be useful in the collection and rapid analysis of weekly feedback that the instructor may
use to inform their course in real time. Here, technology may be particularly useful in
large enrollment courses. However, technologies may also be useful in the large-scale,
detailed analysis of student work (not for grading) but for understanding common student
difficulties around particular content. Technologies may also serve to ascertain a given
instructor's development over time, measuring faculty teaching effectiveness, in part,
based on student learning and development over time.
Finkelstein, A principled way to think about AI in education

12

Are these the roles we want or need for our instructors and students? As we rethink our
educational system, whether to modify our courses modestly to incorporate new technologies, or
grandly to reconsider the structure and objectives of higher education itself, our framing of goals,
roles, and the associated principles can be useful tools.
Administrative roles
While the focus of this piece is on those individuals enacting and engaging in educational
activities, administrators and staff who shape and support those directly involved serve key roles,
as well. These administrative roles include department chairs, deans, those in centers for teaching
and learning and technologies offices, as well as those in higher level leadership such as the
provost and academic leadership cabinet. In addition to considering the principles above and how
these may be supported through administrative actions, these leaders can provide guidance,
resources, and level-setting of expectations. Administrators ought to clearly communicate an
institution’s commitment to and policies around these emerging and rapidly evolving
technologies. These statements include but are not limited to clarifying what constitutes (and
what violates) appropriate use of technology to support campus principles of teaching, learning,
and engagement, and how these practices are evaluated and rewarded. Resourcing the adoption
and effective implementation of these tools will also be necessary. Resources need to include
material support (funding and time) and intellectual (professional development and recognition
of work). Creating faculty learning communities, communities of practice and transformation,
and materials for sharing can advance the capacity of individuals and the institution. Finally,
administrators set expectations both of those enacting and using these tools (faculty and
students), and of the administrations’ expectations of the new workload and practices of
instructors (and students). We ought to take a lesson from COVID-era education where everyone
was asked to do more, simply adding to the workload, which led to faculty overwork and
burnout. Administrators need to consider what to take off educator’s plates, and potentially how
new technologies may ease rather than further add to workload.
Scenarios
A few scenarios provide the application of multiple principles at once to current and potential
coming educational circumstances.
Ubiquitous and university-supported access to generative AI tools: Institutions are now
providing access to subscription service use of AI as part of student enrollment. One can imagine
a world where these new technologies are made available and promoted, but do not come with
requisite preparation, support, or guardrails. Without commensurate training for both students
and educators on how to use these tools effectively, we will miss this moment and may make
things worse. Faculty may ignore these new tools and students use them in unsupervised
fashion, simply plugging essay prompts or homework questions into an AI engine and reporting
the results. In this approach, the enacted goals may devolve to the lowest level, performance vs.
deeper learning, and not developing students’ metacognitive, reflective, and higher order skills
(1.1), and students’ goals (2.1) remain undeveloped.
Alternatively, faculty may seek to regulate technology use in a course, going “back to basics”
and having students take in-class exams and writing essays in blue books, which sends messages
Finkelstein, A principled way to think about AI in education

13

that may counter our overt goals -- e.g. suggesting that professional practice in our disciplines
happens in a 50 to 75 min block, where answers are known in advance, and individuals have to
work in isolation and without the resources available more generally. Thus, AI may not reframe
class but serve to limit current practices (1.3), and undermine the faculty shaping of class (1.1,
1.2). Cheating (2.2, 2.3) may increase, if practices (1.3, 1.4) and goals (1.1) do not adapt. Of
course, there will be individual instructors leaning in and working with students on new class
designs (2.4, 1.5) but if this is not supported and scaled across campus (as per the description of
administrative roles above) it will remain an uneven patchwork. This suite of principles may be
useful in creating stepwise, wide-spread engagement across campus to share collective new
models and approaches.
Supporting basic skills development (helping students "catch up"). An increasing fraction of
students arrive at college without foundational mathematical or writing skills; COVID
exacerbated these preexisting trends. Many have long-considered technology as a mechanism to
support the development of these students. Indeed, there are examples of technology enhanced
skills development (ALEKS, 2025; Carnegie 2025) which go back to the early days of the typing
tutor. Arguably these approaches can support more advanced goals in a course (1.1., 2.1),
utilizing our limited class time on practices that support these goals (1.3,1.4, 2.2, 2.3) and even
afford the time to engage with students on class design (2.4, 1.5). However, we must be
extraordinarily cautious and careful in such approaches. It is easy to falsely attribute a student
poor performance, e.g. on a mathematics placement exam, to student capacity or preparation. In
fact, poor performance is often coupled to lack of preparation and support in learning how to
learn, motivation, trust in education, poverty, or other life circumstances that would not be
addressed by a technology tutor. These students may particularly benefit from enhanced
interaction with other caring humans -- not necessarily technologies. Without attention,
educational innovations tend to exacerbate rather than reduce learning gaps between well-off and
lower-resourced individuals. (Reich, 2020).
Training the AI, developing professional practices. If, in order to learn the basics of a field,
including, but beyond the traditional content, what if we had students developing and training
AIs in the subject domains they are learning? Given that LLMs are customizable, one could
imagine coaching students on how to develop their own agents for further exploring a content
area. Additionally or alternatively, given their capacities for coding and rendering, AIs can be
leveraged by students to represent a content field or domain. Recently we’ve had students (who
have no coding experience) designing computer simulations to represent basic to advance
physics phenomena (BenZion, 2025). In so doing, students must validate AI based approaches
both in terms of the outcomes and scientific accuracy (1.4, 2.4, 2.5). Of course, careful guidance
and scaffolding is necessary, but this approach can support the skills (1.3) that are part of our
professional domains (1.1, 2,1) including: communication, discernment, metacognition, and
trust. In parallel students learn about the strengths and limitations of these new technologies
(2.5).
Addressing Teaching Assistant (TA) and staff shortages. It is entirely likely that there will be
increased teaching demands placed on our faculty and potentially fewer support resources (TAs),
especially at research-based universities. As an alternative to simply increasing number of
courses assigned to individual instructors, we might consider novel approaches through the use
Finkelstein, A principled way to think about AI in education

14

of new technologies. What tasks may be automated? Hopefully, some of the course management
roles (producing a website, attending to the le