Universities have responded to the advent of GenAI with improvised guidance and warnings about misconduct. But the challenge is not to try to exert control over a new technology – it’s deciding how assessment should function in a world where this technology is now readily available. For a credible, meaningful and sustainable approach, I believe we need to move away from this narrow focus on enforcing impossible rules, and open up to a broader rethink of what assessment is for.
Currently, most practices assume authenticity depends on isolating the student from external assistance. But in intellectual and professional work beyond university, that’s not realistic – or desirable. What matters instead is the ability to navigate tools and collaborators to strengthen, rather than dilute, judgement. Here are my five tips for a new approach.
1. Treat AI use as multidimensional
Begin by recognising that “AI use” is not a single behaviour – indeed, AI is not a single tool. Students use AI to brainstorm ideas, improve report structure, generate code, simulate dialogue, summarise and interpret sources, and test arguments. These activities are wildly different in terms of both value and risk, and treating them as equivalent leads to ambiguous policy and inconsistent judgement.
- Assessing students when artificial intelligence is ubiquitous
- Think again: reclaiming our ‘cognitive debt’ from AI
- Assessment tasks that support human skills
Instead, use an approach that distinguishes between forms of assistance that support learning and those that substitute for it. Without that distinction, we risk enforcing rules that are either too blunt to be fair, or too vague to be enforceable.
2. Align assessment to capability
Using AI means we need to look again at the capabilities we’re aligning our curricula to. In some contexts, students’ outcomes should include effective use of AI-enabled tools – but in others, demonstrating autonomous performance is essential. The key issue is not how to monitor AI use, but to specify the capability being valued – and then aligning assessment with it.
3. Reward judgement and reflection
As generative systems become more accessible, the crucial human skill is less often in producing first drafts and more often in exercising judgement. Students need to assess accuracy, identify weak reasoning, compare alternatives and refine imperfect outputs. AI can be valuable partly because it generates variation, offering multiple possible responses that invite evaluation rather than passive acceptance. Used effectively, it can stimulate reflection, challenge assumptions and deepen thinking. Design assessments to reward these higher-order capacities, rather than privileging speed or fluency of production alone.
4. Keep policy simple
For institutional rules to work, they need to be clear, proportionate and workable. Overly complex policies confuse students and are selectively enforced by staff. Place greater emphasis on transparency – that is, mandating declarations or even mandating specific usage. Any policy must also consider equity and workload. Students have unequal access to and familiarity with tools, while staff need time and support to apply rules consistently. Without these conditions, even well-intentioned policies will struggle to function in practice.
5. Redefine independent thinking
Universities have long prized independent thinking, but let’s not confuse independence with isolation. In professional and academic life, capable people routinely use software, data tools and collaboration to extend their work. The relevant question is not whether a student used assistance, but whether they exercised intellectual control over the result. Independent thinking means framing questions, making choices, evaluating evidence and taking responsibility for conclusions. Those qualities remain essential, whether or not AI is involved, and may in fact become more important as machine-generated content becomes more pervasive.
These principles do not mean abandoning standards or accepting uncritical AI use. On the contrary, they place greater emphasis on intellectual accountability. Students should be able to explain, justify and take responsibility for the work they produce, regardless of the tools involved. That requirement is arguably more demanding than traditional notions of authorship, because it shifts attention from origin to understanding.
It also implies that assessment design demands far greater academic imagination and intentionality. As AI lowers the cost of producing plausible text, the differentiating factor in higher education will increasingly be the quality of the questions asked, the rigour of evaluation and the clarity of judgement exercised in response. Those are not easily automated, but they can be either cultivated or neglected depending on how institutions choose to respond.
In conclusion, AI has exposed assumptions that were already overdue for scrutiny. We now face a choice: layer new restrictions on to old models or use this moment to design assessment that is clearer, fairer and better aligned with the skills graduates need. The institutions that respond most successfully are likely to be those that treat AI not as an interruption to endure, but as a prompt to think more carefully about what achievement should mean.
Tom Oliver is lecturer in computer science and engineering at the University of Westminster.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
comment