Dr Mark Carrigan is a Senior Lecturer in Education at the University of Manchester, where he co-leads the Digital Education Manchester group and serves as an AI Fellow at the Institute for Teaching and Learning. His work centres on three interconnected commitments: developing ontological and epistemological frameworks for understanding Large Language Models (LLMs), moving beyond the anthropomorphic and reductive accounts which currently dominate public and academic discourse; examining higher education as a critical site where the social and cultural dynamics of LLMs unfold through practical challenges; and advancing Margaret Archer’s morphogenetic approach as a route to addressing these urgent questions.

Most academics perceive the presence of AI in their classrooms and their student’s work. There is much less agreement however about how we ought to respond to it. Many colleagues feel that defending values which are integral to higher education means keeping AI out of the university. Others agree but see little realistic prospect for how we ought to do this which necessitates some sort of accommodation to AI. Then there are those who feel we must adapt to this technology in order to prepare our students for a future in which it will be ubiquitous. There might even be ways in which we can use the technology to offer more engaging and personalised forms of teaching and learning.

These are conditions which make it extremely difficult to formulate teaching and learning policy.  For policy to be effective it needs to win the support of colleagues in a manner that shape their practice. If we experience purely in terms of compliance, we miss the role it plays in establishing norms and standards. Ideally it creates not just a sense that ‘this is how we do things here’ but rather ‘this is how we should do things here’. In the case of AI it leaves us with a question that we all confront as teachers: what should the role of AI be in teaching and learning? In the absence of agreement we confront students with a cacophony of messages, ranging from condemnations of AI through to claims they must learn how to use it or risk being left behind. This means that students are effectively left to figure it out for themselves when presented with inconsistent guidance.

Why assessment guidance needs dialogue, not just compliance

This is the context in which the AI in Teaching and Learning policy was developed, before being passed by Senate in April. It brought together a diverse group, including academic and professional services colleagues alongside student representatives, with an equally diverse range of perspectives on the role of AI in teaching and learning. The conclusion we reached was that we urgently needed to provide students with much greater clarity about expectations concerning appropriate use of AI in their assessments. We had struggled to offer a clear institutional position since the launch of ChatGPT in November 2022 and this was increasingly untenable. However it was also clear there could be no final answer to these questions because the landscape is changing so rapidly and levels of AI literacy vary so widely amongst students and colleagues. Furthermore, in a large multidisciplinary university there will inevitably be different disciplinary considerations which means norms and standards will vary across our faculties. Attempting to be too prescriptive would inevitably be counterproductive but equally there was an urgent need for a path forward.

The new policy does this by asking unit leads to classify their assignments into one of four categories, ranging from AI being entirely prohibited through to AI being required to complete the assignment. The expectation is that most assignments will fall into the second and third categories which allow minimal AI use (such as for copy editing) or specific uses of AI which will be signalled to students in guidance. It departs from the traffic light systems, ranging from prohibition (red) through context-specific (amber) and encouraged (green), which have become commonplace through the addition of the extra category. An obvious risk here is that this additional category risks creating ambiguity. At what point does AI Minimal become AI Permitted?

From traffic lights to reflective judgement

The advantage of this approach is that it encourages us to draw these distinctions. It invites reflection by unit leads and conversation within teaching teams and programmes about these boundaries and what they mean for assessment. The problem with the traffic light system tends to be that most assignments get categorised as amber, which means the question of clarity gets displaced. If a unit lead categories something as amber because it seems like the safe option, it is only a limited improvement for students because the context-specific guidance will still be lacking. In contrast our approach builds towards more fluent and fine-grained judgements because these are exactly what we need in order to provide our students with clearer expectations about the role of AI in teaching and learning.

It is an ‘approach’ though, not a solution. AI in assessment is a classic example of a wicked problem. There are only better or worse options, involving trade offs. If we imagine that a classification system is going to resolve the problem so that we can return to the status quo ex ante, we are failing to look at the sheer scale of the technological challenges which are still emerging. From coding agents through to the rapid growth of wearable computing, from AI companions through to AI labs ‘enshittifying’ their models, the emerging challenges are even more consequential than the ones we’ve faced since the launch of OpenAI’s ChatGPT in November 2022. If we can’t solve these problems, we can at least make sure we are moving in the right direction. Initiating a deeper conversation about what AI means for teaching and learning is a necessary condition for that forward movement. Nearly two years ago The Russell Group principles called for “Engagement and dialogue between academic staff and students” in order to “establish a shared understanding of the appropriate use of generative AI tools”. It was essential that “this dialogue is regular and ongoing will be vital given the pace at which generative AI is evolving”.

The new policy creates the structural conditions to support such a dialogue. Firstly, by asking unit leads to reflect on their assignments in order to identify the expectations which should be communicated to students. Not only does this encourage dialogue with students, it invites dialogue within the teaching team where these expectations might not already be clear. Secondly, by establishing an annual review of a policy which will need to be iterated to keep pace with technological developments. In itself this is a governance mechanism rather than the thick dialogue suggested by the Russell Group principle. But in inviting conversations about the policy, what is working and what is not, it creates the condition in which that more substantive dialogue is valued and encouraged. The policy will not bring that about in itself but it can contribute to making these conversations a routine feature of our work together within the university.

Further reading:

• Responsible AI in Learning: A Student-Informed Approach
• Assessment and AI
• An Essay in Seconds: What AI Reveals About Broken Assessments and the Need for Pedagogical Reimagining

Related