Implementing a collaborative program-wide approach to redeveloping assessment in response to generative artificial intelligence (GenAI)

Synthesis: Nicola-Richmond et al. (2026) report a combined ethnographic and interpretive phenomenological study of a collaborative program-wide project to redevelop Assessment in response to Generative AI in a School of Health at a large Australian university. Thirteen participants — academic team members (n = 4), project leads (n = 3) and researchers (n = 6) — contributed 8 project meetings, 3 focus groups and 12 reflective diary entries, analyzed thematically. The study documents what a program-wide approach actually feels like to the people doing the work: low confidence and strong negative emotions early on, views that evolve from avoidance toward ethical integration, a persistent gap between principles and concrete task-level practice, and five success features — a range of expert stakeholders working collaboratively, openness and curiosity, professional development, protected time for sharing and reflection, and industry input. The authors' central claims for practice: designing quality assessment is the same challenge it has always been but harder, and GenAI is also a "trojan mouse" opportunity to fix long-standing assessment weaknesses.

Key Findings

• Confidence and aptitude were the entry barrier, not resistance. Most participants described a lack of confidence and skills in what to do about changing assessment — one academic team member reported "a lot of uncertainty around my own ability to be up to scratch on what I need to be up to scratch on."
• Emotions were strong and mostly negative. Participants described feeling "daunted," "nervous" and "quite freaked out" in response to student GenAI use and to the work of changing assessment; the authors found these emotions central to the experience of assessment change, and the program-wide approach helped manage them.
• Views evolved measurably over the project. Participants shifted from "ensuring students are avoiding GenAI use" toward "how can GenAI be used ethically and effectively to support learning," and from "I'll just build it into every assessment" to a more nuanced position — a shift one project lead called "quite fascinating" to watch over 12 months.
• University guidance lagged practice. Participants repeatedly asked for clearer institutional guidance on acceptable GenAI use for staff and students, particularly on Academic Integrity; one academic reported avoiding the question entirely: "I'm just sort of avoiding it [dealing with student use of GenAI], because I'm not clear with what we're allowing and not."
• The work is heavier than it looks. Redesigning an assessment meant rewriting the assignment, redoing the rubric, going to the assessment panel, responding to feedback and updating the LMS — a sequence academics described as a significant time burden.
• Options for assuring learning are genuinely narrow. Beyond oral and performative modes — hard to scale to large cohorts and online programs — participants found few designs that minimize GenAI use; one project lead had expected "a whole range of solutions" and found that "this doesn't seem the case."
• Program-wide was chosen pragmatically, on resourcing grounds. Unit-by-unit redesign was unviable on time and cost; the program-wide lens also gave better visibility of scaffolding across a qualification than unit-level approaches, and let teams scaffold "from less sophisticated to more sophisticated use of [GenAI] over the program."
• Collaboration was the core affordance. Assessment redesign in the GenAI era exceeds any single academic's capability: participants valued learning designers, assessment design experts, GenAI expertise and — increasingly — industry voices, with one project lead wishing for "an actual tech head" with coding knowledge too.
• Concrete output to date. The project changed 18 assessment tasks to assure learning (17 more proposed) and modified 4 to incorporate GenAI use (6 more proposed), with seven additional programs joining in 2025.

Five features participants identified for successful implementation

1. A range of expert stakeholders working collaboratively — "having investment from all of the people… the people who are the experts in GenAI at the moment, people who look at assessment design at university level," creating "that really collaborative experience for everybody across all of the units and all of the year levels."
2. Being open and curious — "it's been important… to be curious and open and a bit vulnerable and a bit uncertain. And that's all being welcomed and okay. Everybody getting a seat at the table."
3. Professional development and ongoing support — participants wanted "equal access to PD [professional development] or opportunities for mentoring," provided up front and available throughout.
4. Protected time to share and reflect — "One of the really key things was letting everyone have a reaction and explicitly stating that all views are okay," so hesitation, experimentation and inexperience were all treated as legitimate.
5. Industry input — to check whether redesigned tasks would disadvantage graduates: "industry will be saying, 'Why are you not using [GenAI], doing this?'"

Four discussion points for assessment design in an age of GenAI

• Designing quality assessment is the same challenge, but harder. The classic difficulties (time burden, expertise gaps, sustainability) all reappear, now compounded by staff lacking GenAI skills and needing clearer guidance. One difference from earlier work: participants were motivated to engage — GenAI's arrival created "a focus on assessment design that likely should have already been occurring."
• GenAI is a "trojan mouse" opportunity. Just as learning management systems in the 1990s opened conversations about learning and teaching through technology conversations, GenAI is opening "time and investment for better conversations about assessment" — and pushing institutions from fragmented unit-by-unit assessment toward the shared assessment cultures and better learning scaffolding that program-wide approaches enable.
• A program-wide approach is a pragmatic choice, not the only one. The authors are explicit that a programmatic approach (van der Vleuten et al.) or unit-based redesign could be preferable in some contexts; the program-wide lens was chosen for timely, manageable change under real resourcing constraints.
• Moving from principles to practice is the hard part. The gap between conceptual principles and feasible structural changes to individual tasks was the most-cited difficulty; leaders felt pressure to be "the expert" while acknowledging they were "nowhere near an expert."

What this means for practice

• Instructors. Expect the emotional load and normalize it: participants reported feeling "daunted" and "quite freaked out," and a collaborative program-wide setting was what made those feelings manageable — run assessment-change work in teams, not alone, and set explicit task-level rules when institutional guidance on acceptable GenAI use is still unclear rather than avoiding the question.
• Faculty developers. Fund the full redesign loop, not just the writing: participants counted rubric rework, assessment-panel review, feedback cycles and LMS updates as part of every single task change — resource the sequence or it stalls.
• Faculty developers. Assemble the whole village deliberately and invest up front: assessment design, GenAI and technical expertise, learning design, subject knowledge, program mapping and industry voices, with equal access to professional development and mentoring sustained throughout.
• Administrators. Protect the meeting time where "all views are okay": the sharing-and-reflection slots were what let views evolve from avoidance toward ethical integration, which is where the redesign gains came from.
• Administrators. Scaffold GenAI use deliberately across a qualification — from less to more sophisticated use over the program — rather than leaving each unit to decide independently, and plan for scale limits: few assurance-of-learning designs scale beyond small, on-campus cohorts, so budget for oral or performative assessment where cohorts allow and accept that online and large-cohort units need a different answer.

Limitations

• Single-institution study in one School of Health at a large Australian university (ethics approval HEAG-H 209_2023), so findings may not transfer to other disciplines or institutional contexts.
• Small purposive sample of 13 participants (4 academics, 3 project leads, 6 researchers), all involved in the project being studied, with the authors acknowledging interpretive bias and using reflexivity plus an independent research assistant to counter it.
• The research team members were also project participants and leads, so the evaluation of the approach is partly self-evaluation despite the triangulation of meeting recordings, focus groups and diaries.
• The paper reports only the first iteration of the project (2024, three programs) as preliminary evidence; later iterations and the seven programs that joined in 2025 are not yet analyzed here.
• Data capture spans a rapidly shifting technology and policy environment; participants' views and the institutional guidance they were waiting for both moved during the study.

Citation

Nicola-Richmond, K., Dawson, P., Lawson, J., Fay, P., Parson, J., Macfarlane, S., Riddell, S., & Finneran, R. (2026). Implementing a collaborative program-wide approach to redeveloping assessment in response to generative artificial intelligence (GenAI). Assessment & Evaluation in Higher Education. (CC BY-NC-ND 4.0.)