December HE AI Community Meetup - Artificial intelligence

This community meetup brought together colleagues from across higher education to share live practice, frustrations, experiments and unresolved questions about using AI in teaching, learning and assessment. The discussion was wide-ranging, grounded in institutional experience, and focused firmly on what is happening now.

What emerged was not a set of answers, but a shared picture of where the sector currently is, what is working, what is proving difficult, and where attention is shifting next.

AI chatbots as teaching tools, what is working and what is not

A significant part of the discussion focused on student facing AI tools embedded within learning platforms, particularly role play and Socratic style chatbots.

Several institutions are actively using AI Conversation tools within Blackboard Learn Ultra and Coursera. These include structured role play scenarios and guided questioning aligned to module content. The value was not framed as the AI itself, but as what it enables pedagogically.

Participants described these tools as a way of bringing otherwise passive online learning environments to life. They introduce active learning into spaces often dominated by static content, prompting learners to respond, reflect and articulate their thinking rather than simply consume information.

Student feedback from pilots introduced in 2024 reflected this mixed but generally positive experience. Learners reported that conversational formats helped them think more deeply, expand their answers and recognise the complexity of topics such as sustainability and technology in business. Some found being questioned in response to their own answers particularly effective. Others were less enthusiastic and found the interactions repetitive or dull. This variation was seen as expected rather than problematic.

A recurring caution was that some chatbots do not naturally conclude an interaction. Without careful instructional design, students can find themselves in open ended conversations that add little educational value. Clear purpose, boundaries and framing matter.

There was also discussion about configuration. Many tools operate as walled gardens, drawing only on course materials and approved contextual documents rather than the open web. For formal education settings, this was broadly viewed as a strength.

Scaling feedback and support for distance learners

For providers with large numbers of distance and flexible learners, AI supported dialogue tools were seen as a potential response to long standing feedback challenges. Students consistently ask for more opportunities for interaction and formative feedback, but delivering this at scale has always been difficult.

Early experience suggested that learners do not see these tools as a replacement for staff engagement. Instead, they value them as a complement to peer feedback and tutor input. When used carefully, they can extend support without undermining the role of educators.

Assessment, policy confusion and the student experience

One of the strongest themes was student confusion caused by inconsistent staff approaches to AI.

Many institutions have adopted traffic light or lane-based frameworks for AI use in assessment. While intended to provide clarity, these are often interpreted through individual staff preferences rather than shared understanding. Students are left trying to work out what a particular tutor likes or dislikes, which was widely seen as unacceptable.

This confusion is particularly acute for students using assistive technologies. Learners prescribed tools such as Grammarly are unsure whether they are allowed to use them when an assessment is labelled as AI designed out. These students are not trying to game the system, they are trying to comply, and anxiety levels are high.

Participants were clear that students should not be navigating personal staff views. A coherent institution-level position is essential, even where local interpretation remains necessary. Several contributors described the same debates replaying across departments, faculties and governance structures.

There was also frustration around AI embedded in core systems. Library databases and discovery tools increasingly include AI driven summarisation and semantic search by default. In some cases, students were told not to use AI during library skills sessions, even though it is built into the platforms themselves. This creates contradictions that are impossible for learners to resolve.

Institutional consistency versus disciplinary nuance

Alongside calls for institutional clarity, there was recognition that disciplines differ. Assessment cultures, ways of knowing and acceptable practices are not uniform across subjects, and overly rigid policies risk flattening these differences.

One pragmatic response shared was the introduction of a mandatory generative AI section in module and programme handbooks. This does not prescribe practice but requires teams to explicitly discuss and articulate expectations. Simply surfacing the issue was seen as meaningful progress.

The balance between consistent student experience and disciplinary autonomy remains unresolved. What the group agreed on was that ambiguity and silence are worse than imperfect clarity.

Rethinking assessment in the age of AI

AI continues to expose weaknesses in assessment design. There was broad agreement that if an AI can complete an assignment without detection, the real question is what was being assessed in the first place.

Rather than using AI to audit assessment for fault, some institutions are using it as a prompt for reflective review. Teams are revisiting alignment with learning outcomes, the evidence tasks produce, and whether assessments genuinely test understanding rather than recall.

Authentic assessment, grounded in application, judgement and context, was repeatedly emphasised. At the same time, participants acknowledged that recall under pressure still matters in some professional domains. The challenge is not removing knowledge but being explicit about what kind of knowledge and capability is being assessed.

Evaluating AI output, learning gains and trade offs

The group explored the growing practice of asking students to evaluate AI-generated output. This approach has clear benefits for developing critical thinking and AI literacy, but it also raises questions.

Concerns were raised about opportunity cost. Time spent critiquing AI output is time not spent engaging directly with disciplinary literature. Others questioned whether evaluating AI responses is pedagogically sound, given that AI systems are not accountable in the way human tutors are.

Despite this, there was strong consensus that the ability to critically evaluate AI output is now a core skill. AI mediated information is pervasive, and students need to recognise bias, misinformation, weak sourcing and inappropriate framing.

Examples shared included assessing prompt quality, source credibility, potential bias, language and structure, and relevance to disciplinary context. A caution was raised that students may also use AI to generate the evaluation itself, which needs to be considered in task design.

Everyday AI and the myth of choice

A quieter but important thread was the normalisation of AI in everyday tools. Writing support, dictation, summarisation and editing features are now embedded in common software such as Word. For many staff, this has been genuinely transformative, particularly for accessibility and productivity.

The idea that AI is something people can simply choose not to use was strongly challenged. In many cases, AI is already present by default. The question is no longer whether to allow AI, but how to work responsibly with tools that are already embedded in institutional systems.

Priorities for the year ahead

Looking ahead, several shared priorities emerged.

Some institutions are focusing first on academic integrity processes, aiming to address tensions between detection tools, student accounts and fairness before tackling larger assessment redesign.

Others are developing AI literacy frameworks, exploring how existing models apply to their context and where adaptation is needed. A recurring gap identified was the need for space for ethical judgement and decision-making, not just technical skills.

Across all priorities, there was a strong call for time and space for staff to talk. Not just training sessions, but opportunities for discussion, disagreement and shared sense making.

The next community meetup, at 3.30 pm on 13^th January, will continue these conversations into the new year. Whether you attend every session or drop in occasionally, the value lies in sharing practice honestly, naming concerns, and working through complexity together.

Links shared during the call

Find out more by visiting our Artificial Intelligence page to explore publications and resources, learn more about our communities and sign up for our AI Literacy training.

For regular updates from the team sign up to our mailing list.

Get in touch with the team directly at AI@jisc.ac.uk