Seventy delegates from nineteen institutions convened at The Pā for a full day on the future of learning and assessment in Aotearoa. Seven of New Zealand's eight universities were represented. The programme opened with a pōwhiri, closed with a whakamutunga, and placed three student representatives ahead of every institutional session.
Across the day a consistent position emerged. Detection has reached the limit of its usefulness, assurance of learning is the successor problem, and the assessment volume accumulated since the pandemic is now the primary constraint on solving it. This report is organised into six themes.
The Wharenui at The Pā, mid morning.
How this report is organised
The six themes
Each theme carries its own colour through the rest of the report, along with the sessions that speak to it. Every session is recorded and linked from the theme it belongs to.
The student experience of assessment
Three student representatives opened the programme, ahead of every institutional session. Their account of assessment as it is actually experienced set the terms of reference for the day, and was cited back by speakers in every session that followed.
- Assessment load is cumulative, and no one designing a course can see the total. Most students the Auckland University Students' Association represents work 20 to 30 hours a week on top of their study, and deadlines set independently across four papers converge without coordination. Matthew Lee's suggestion was practical and easy to act on, to release a past exemplar so students can see what a finished piece looks like, since a rubric describes the standard without showing it. His case for it was a case for trusting students, that an exemplar is used for inspiration far more often than it is copied. Matthew Lee
- Students locate the integrity boundary at the substitution of their own reasoning. Seamus Lohrey put the line where AI replaces your own critical thinking, and was candid that knowing the line does not hold students to it. His explanation is that nobody has taught them the skill, in the way the library once taught research method. What he asked the room for is that AI use be taught as explicitly as research method now is, so that the boundary students already recognise becomes one they are equipped to hold. Seamus Lohrey
- Prohibition without a stated rationale is the position least likely to survive. AI is barred outright across several law papers while the firms those papers prepare students for are investing heavily in it. Lohrey's answer was not permission but explanation, telling a class why a particular task exists and what thinking it is meant to build. He credited Waikato's own approach and located the gap in staff uptake rather than in policy, which is where the work now sits.
- Identity verification distributes its burden unevenly. For the 60% of Massey's cohort studying at a distance, remote verification means a room scan, a desk scan and photo identification that distance students are rarely issued in the first place. Ayla Brook made the case for both things being true at once, that verification is entirely legitimate and that it asks more of some students than others. She was honest that the answer is not an obvious one. What she pointed to was notice far enough ahead for students to prepare the space and the documents, and thought given to students who are working from a dining table or a car, including the study rooms some libraries will book out. Ayla Brook
- The assessments students recalled required performing the discipline rather than describing it. Acting on a family law brief with a hundred-page client file, drafting a constitution for an invented state, taking an indigenous rights intervention to a mock United Nations session. Lee put the principle behind all three in a sentence, that students study in order to do the thing they are studying, so an assessment may as well ask them to do it. Seamus Lohrey added the marker's half of it, that the assessments worth setting are the ones where any well reasoned answer can earn full marks.
I think we can probably trust our students a bit more, to not copy that exemplar, but actually use it for inspiration.
Matthew Lee · Undergraduate Education Vice-President, Auckland University Students' AssociationAssessment volume and the case for reduction
Assessment volume surfaced in every session, raised independently by students, practitioners and Deputy Vice-Chancellors. Two waves of accumulation have occurred since the pandemic, and the room was close to unanimous that reduction is now a precondition for reform rather than a consequence of it.
- The arithmetic is institutional rather than individual. Professor Jennifer Milam described some papers as carrying 10 to 40 assessment items, and put the total a student across four papers can be facing at 40 to 120. No individual course designer has sight of that total. A student leader had told her the day before that the load was beyond what could be completed, and that some courses set assessments daily without teaching daily. That arithmetic is what the policy work now under way at Waikato is designed to answer. Professor Jennifer Milam
- Two waves, differently motivated, neither designed with the whole load in view. Assessment was added coming out of the pandemic as a mechanism for attendance and engagement, on the premise that students will not do work they are not marked on. It was added a second time as a control on generative AI, to build tasks that could not be completed with it. Both waves were rational one course at a time and neither was visible at the level of a student's whole programme.
- Waikato proposes reducing to three. The proposed policy would take assessment down to three items per paper, phased over two years and commencing at 100 level, alongside a university-wide conversation about learning outcomes and how they relate to the graduate outcomes each programme is built for. Milam was clear that the reduction is the means rather than the end. The question Waikato will ask itself two years from now is whether more space for learning has appeared inside its papers.
- Structural custom is in scope. Jean Jacoby asked whether a bachelor's degree really requires 24 discrete slots of 15 credits, and made the case that oral assessment and the rest belong inside courses rather than on top of them. Her point was that custom and practice have held these shapes in place for a long time without being examined, and that the examination is now overdue. She put it to the room as the conversation universities should be having next. Jean Jacoby
- A programme view relieves the individual course, and accreditation sets the limit. Stephanie Reid described Auckland moving to a whole-of-programme approach, in which a course may legitimately carry no assessment at all because assurance of learning happens at identified points elsewhere in the qualification. She put it to academics as a prompt rather than a rule, that the question is where across a programme the assurance points belong. She was equally clear that external accrediting bodies set out what they need, and that a programme map has to be built around them. Stephanie Reid
Associate Professor Kirsten Petrie closed the institutional panel by applying the standard reflexively. Academics did not enjoy being assessed by the Performance-Based Research Fund once every six years, and yet assessment at something closer to weekly is routinely asked of students. She offered it as a provocation rather than a finding, and ended the session on it.
We over assess and under teach.
Seamus Lohrey · President, Waikato Students' UnionTrust as the foundation of assessment design
Professor Kathryn Sutherland reframed the question the sector has been asking since 2022. Institutional attention has concentrated almost entirely on whether students can be trusted. Her session asked what evidence exists that students trust us, which teaching practices generate it, and what the data says when you go and measure it.
Trust Moves · four dimensions
- Cognition. Demonstrated command of the field, and the presence of the teacher's own research in the room. This is the position most students begin from.
- Affect. Demonstrated care for the individual student. This is the dimension students named most frequently and without prompting.
- Values. Integrity in conduct, and fairness in marking, enacted rather than asserted.
- Identity. Acknowledgement that the room is not composed of people like the teacher, and design that follows from it.
Built with Rachel Forsyth and Peter Felten from interviews with 29 experienced academics across four countries, and tested with nearly 500 students across four universities and three countries. Students rated all four as mattering. Asked in open response to name a lecturer they trust and say what that person does, they wrote about being known and cared for more often than about expertise, which puts affect among the most available dimensions to any teacher on any Monday.
The interventions the framework implies are small and immediately available. Learning students' names. Releasing the practice question ahead of the assessed one. Co-designing rubrics with the cohort. Making the purpose of an assessment explicit, which is the operating principle of the Transparency in Learning and Teaching framework she cited.
One carried a result. An interviewed colleague replaced a single high-stakes component with around five short preparation exercises a week, marked zero, one or two, where a student reaching half the available points takes full marks for that component. Students can skip a week when another course is heavy, which makes the scheme a workload valve as well as a practice run. His reported effect was that he now receives very few extension requests, and that students no longer need to find answers anywhere other than in the attempt itself.
Sutherland identified a disclosure asymmetry that two later speakers returned to unprompted. Disclosure of AI use is required of students continuously and volunteered by academics rarely. The phrase she borrowed for the environment this produces, an architecture of fear, entered the vocabulary of the day.
Students don't come to university to cheat.
Professor Kathryn Sutherland · Centre for Academic Development, Victoria University of WellingtonTwo lanes, and the programme as the unit of change
The two-lane model is now the settled direction across the New Zealand sector. The substantive question in the room was the unit at which it operates, and every institution attempting it at course level reported the same constraint.
- University of Auckland adopts the two-lane nomenclature in 2027, and is planning for redesign to continue well past the date rather than finish on it. The reframe underneath it is the part Stephanie Reid emphasised, that the question has moved from whether a student used AI to how they used it. The support around it is learning design capacity and workshops for academics, with process over product as the organising principle. Stephanie Reid
- Massey University secured academic board approval mid-year for full implementation in 2028, an 18-month rollout that Jean Jacoby characterised as tight. Her reason is that two lanes only work properly on a programme view, which means looking across an entire qualification and rethinking the purpose and placement of every assessment in it. That is the work now under way, and the timeline reflects how much of it there is. Jean Jacoby
- The AI use framework it replaces was retired for a specific reason. Massey built an intermediate framework on the AI Assessment Scale, intended to give students and staff a common vocabulary for describing AI use, and always expected to progress to two lanes in time. What the framework surfaced along the way is the most useful finding of the session, that categories meant for describing AI were being used to police the amount of it. Jean Jacoby described the move to two lanes as a progression along a path Massey was on from the beginning. Jean Jacoby
- University of Canterbury reached a milestone of 10,000 online learners across a four-year build, and Professor Mick Grimley attributes what works to deliberate design, a dedicated learning design team, cohort personas and analytics read while students are still learning rather than afterwards. Assurance is the frontier he is most interested in next, and Canterbury is looking hardest at oral assessment as the route to it, with scaling the open question. Professor Mick Grimley
- Toi Ohomai Institute of Technology works with whole teaching teams rather than individual educators, which Josh Burrell described as their preferred approach because it produces a consistent one across a broad vocational portfolio. The structural advantage he named is that much of the teaching already happens in labs, workshops and kitchens where learning can be observed directly, and the work now is carrying that into the higher levels of the portfolio. His framing was that confidence across a team sits on a bell curve, and the people who most need bringing along are not the ones who turn up to conferences. They can sit in the back seat for a while, he said, as long as they are on the bus. Josh Burrell
Stephanie Reid described leading the change as a tricky dance and at times a clumsy one, and said the conversation is now catching fire among the academics her team works with, with students next. Professor Jason Stephens made the structural case from the opposite end of the institution. An individual course absorbs assurance work it should not have to carry wherever the programme holds no view of the whole.
It is not whether students have used AI. It is how they are using it.
Stephanie Reid · Associate Director, Learning and Teaching Design, University of AucklandFrom detection to assurance of learning
Four practitioners brought redesigns they had already run, with the results. Professor Jason Stephens set the frame with three years of institutional caseload data from the University of Auckland, across which plagiarism cases halved while generative AI cases rose roughly threefold.
- A midterm moved into a platform that records the writing, with the marking done first. Teaching a stage-one course of 391 students, Stephens spent the first half of the semester running low-stakes weekly writing tasks so the cohort was fluent in the tool before anything rested on it. The 90-minute midterm followed, with 367 submissions marked in full before anyone looked at the process data. Twenty-five submissions then surfaced as unclear or only moderately assured, about one in fifteen, and all but one student acknowledged the breach when asked. He recorded them as poor academic practice rather than misconduct, and described it as the least contentious possible way of handling an integrity case. Professor Jason Stephens
- Remote invigilation was tested by asking staff to defeat it, and they did. Professor Mick Grimley set his own staff a task under invigilation software and two cameras, and instructed them to get around it however they could. They used earphones with a phone on the floor and macros written on the machine, and he was struck by how hard it was to spot any of it on camera. His conclusion was that invigilated examinations at a distance are difficult to envisage, which is why Canterbury is looking at oral assessment instead. Professor Mick Grimley
- Assessment rebuilt in thirds, with the caseload and the grades both measured. At Massey University College at least half of all assessment moved to supervised conditions, and the take-home half was rebuilt so that a research portfolio, an oral kōrero and a written submission each carry roughly a third. A student cannot pass on the written artefact alone. Cases raised by educators fell to around a quarter of the previous semester, and the prediction the team expected to be proved right, that grades would drop, did not happen. Averages held, and students following the research process came out slightly ahead. Erica Lee and Regan O'Malley
- Making the notes an artefact showed where synthesis was failing, not where integrity was. Megan Clune kept the literature review and rebuilt the path to it, adding a thematic note-taking matrix, article uploads and an optional draft paragraph returned with feedback, which about 15 of her 20 students took up. Reading the matrix against the paragraph produced the finding she did not expect. Students had the right ideas in their notes and those ideas were not arriving in their writing, which let her give specific feedback on the gap. The one student she looked at more closely was the one whose work was weaker than she expected, and the data told her how to help. Megan Clune
- The clearest evidence of copied work turned out to be the arithmetic of typing. Clare Crawshaw's cohorts average 11 hours on a 1,500-word report, much of it in class time with the work open in front of them. A student transcribing something written elsewhere takes about an hour and a quarter, and the record shows a document filling from the first word to the last with almost no revision. Her submission rate across six or seven assessments has not fallen below 97%, and 88% of her students rate the experience good to excellent, which she puts down to setting the cohort up from the beginning as much as to the platform itself. Clare Crawshaw
I'm not saying you cheated. I just don't have assurance of your learning.
Professor Jason Stephens · Academic Head, School of Learning, Development and Professional Practice, University of AucklandWatch this theme
The purpose of the university
Beneath the sessions on policy and the sessions on practice, a prior question recurred. Several speakers arrived at it independently and none treated it rhetorically, which made it the most useful conversation of the day to have started.
- What a university is for once text is no longer scarce. Jean Jacoby asked what generative AI means for knowledge itself, and what follows for an institution whose currency has been the production of written work. Her method for getting at it is to turn the light on everything the university takes for granted and ask why it is done that way. The support question she raised alongside it is the practical one, that few of Massey's staff came to teaching as trained teachers, and they are now being asked to teach students how to think about their own thinking. Jean Jacoby
- What a student gets here that they could not get elsewhere. Stephanie Reid put the question in its hardest form, against an online provider with an AI teacher. Her hope for five years from now is that the sector is well into a mature conversation about what a university is for, and further down that track than the reactive work of the last three years has allowed. Stephanie Reid
- The reciprocal part of teaching has been displaced, and there is a test for getting it back. Jennifer Milam identified a quieter cost, that what educators enjoy most is learning from their students, and that the administration of assessment has crowded it out. Her measure for whether Waikato's reset has worked is not a count of assessments but whether that reciprocity has room to happen again inside papers, in partnership with students rather than in administration around them. Professor Jennifer Milam
- A vendor that built detection first and threw it away. The first thing Cadmus built, in a year of incubation with Melbourne, was a contract cheating detector, and the students it was applied to told the team plainly that they did not want it. Two years in the team stopped and listened properly, and what they heard was that students felt unsupported in their learning and that this was driving the misconduct downstream. They rebuilt from scratch around supporting learning instead. Herk Kailis told the forum that misconduct rates fell by 90% once the design question changed. Herk Kailis
We will protect the human where it's most valuable.
Jean Jacoby · Director of Education Futures, Massey UniversityEvidence
What the redesigns produced
Four practitioners brought redesigns they had already run, with the results. These are their own figures, from their own cohorts.
- 1 A research portfolio, submitted first
- 2 An oral kōrero, before the final draft
- 3 The written submission
Massey University College. A student cannot pass on the written submission alone. The portfolio comes first and earns formative feedback, and most lecturers now run the kōrero before the final submission so corrections can be made while they still count. It is a conversation about the research process rather than an interrogation of the text, it runs 10 to 15 minutes from a one-page outline, and pronunciation is explicitly excluded from the rubric.
Megan Clune retained the literature review as the assessed task and rebuilt the path to it, introducing a thematic note-taking matrix and an optional draft paragraph returned with formative feedback. About 15 of her 20 students took the draft up, and reading the matrix against the paragraph showed her where synthesis was breaking down.
Bronwyn Alton rebuilt a group assessment around a shared checklist, a timeline and an agreed division of work, so contribution is visible to every member of the team and she can intervene early. Student satisfaction with the assessment rose, and the divide-and-reassemble pattern that group work usually produces gave way to teams working through the process together.
Clare Crawshaw runs a live class exercise in week two in which students paste text from Wikipedia, then from ChatGPT, and observe the platform register both. She then shows them the educator view in full.
Toi Ohomai Institute of Technology · Level 5 Business Diploma
