Choices, not policing: reshaping assessment for the AI era

Advance HE and Inspera bring senior academic leaders together to consider trust, the 2028 regulatory horizon and the future of teaching and learning. 

More than 50 senior academic leaders from across the UK gathered in Manchester last for The Assessment Forum, a joint event from Advance HE and Inspera exploring how institutions can build secure, credible assessment in an era of generative AI. Rather than a conventional lecture format, the day was built around three panel provocations, each followed by extended table discussion, with delegates capturing their own observations via Padlet. A short paper drawing on the day’s discussions will follow.

As Professor Mark O’Hara, Senior Consultant (Education) at Advance HE, told delegates in his opening remarks, the sector is meeting “at a period of profound transition.” The rapid, near universal adoption of generative AI by students is reshaping how institutions think about assessment, assurance and trust. The task of the day, he said, was to step back from the noise and think strategically about what secure, credible and future proof assessment should look like. 

Panel 1: The cost of trust 

There was strong consensus that a retreat to closed book exams is not the answer. Professor Scott argued that institutions must return to first principles and ask what assessment is really trying to achieve, respecting that different disciplines call for different approaches. Professor Williams, speaking as a neurodivergent early adopter of AI, was clear that AI has not broken assessment: it has exposed practices already struggling to provide authentic evidence of learning. She described a staged model at Queen’s, with clear points where AI use is encouraged, controlled or restricted. 

Professor Jessop spoke about the shift to programme level assessment at Bristol, reducing the overall assessment burden while rebuilding more engaging, human centred teaching. Dr Alston offered a practical framework: institutions should ask what capabilities are being assessed, what the consequences of getting it wrong are, and what the most authentic way is for students to demonstrate their ability. Security, he argued, should be proportionate to the judgement being made, with educational design doing the heavy lifting rather than blanket restriction. 

Panel 2: The 2028 regulatory horizon

Ms Coward warned against a reflexive return to on premise, invigilated exams, describing this as “an 1826 answer to a 2026 problem.” She cautioned that if the sector does not respond coherently, the Office for Students could be pushed towards new conditions of registration requiring auditable AI assessment policies, a shift that would constrain genuine innovation in teaching. 

Professor Grogan agreed that institutional responses to date, ranging from outright prohibition to informal traffic light guidance, will not be sufficient on their own. He argued that as AI makes the finished essay “cheap,” the real currency becomes the verifiable human process behind it: judgement, authorship and authenticity. Both panellists were united in urging the sector to regulate itself, in genuine partnership with students and employers, before external regulation is imposed upon it. 

Panel 3: Teaching, learning and the digital education horizon

The final panel, chaired by Vic Stephenson, Senior Consultant (Education) at Advance HE, brought together Professor Alistair Robertson (Queen Mary University of London), Professor Anne Murphy (University of Portsmouth) and Dr Miri Firth (University of Manchester) to discuss embedding AI literacy across institutions. 

Professor Robertson described Queen Mary’s approach to critical AI literacy, contextualised discipline by discipline, alongside a red, amber, green framework in which “green” activity formally assesses students’ AI literacy within their subject. Professor Murphy stressed that colleagues first need support to articulate the core knowledge and skills of their discipline before AI can be meaningfully built around it, and argued for equitable, institution wide access to tools alongside firmer accountability for take up. Dr Firth reframed the challenge around graduate competition: students compete not against AI but against peers who use it more effectively. She highlighted innovations such as interactive rubrics and the importance of student trust in institutional intentions around permitted AI use. 

Conclusions 

Bringing the day together, Vic Stephenson, Advance HE Senior Consultant Educational Excellence, reflected that the value of the forum lay less in any single provocation than in what she called “the wisdom in the room”: the chance for institutions navigating the same choppy waters to compare notes and learn from one another, rather than treating assessment as something to police. For her, that spirit of shared, sector led problem solving is exactly the role Advance HE exists to convene. She invited delegates to share what they felt would be most useful for Advance HE to take forward in supporting the sector in the effective, inclusive and ethical use of AI in higher education. 

Ishan Kolhatkar of Inspera offered a similar note of cautious optimism, observing that four years into the mass availability of generative AI, what increasingly sets people apart is not access to the technology but the human judgement to interrogate, question and challenge it. Building that confidence, he suggested, will be the shared task of institutions and technology partners alike in the years ahead.