ÌìÃÀ´«Ã½

The ‘Advancing Feedback Practice with Generative AI: Pedagogical Challenges and Opportunities’ conference, hosted by the Centre for Higher Education Research and Scholarship (CHERS) at ÌìÃÀ´«Ã½ on 23 April 2026, recently brought together educators, researchers, ÌìÃÀ´«Ã½ and learning designers.

The conference formed part of Imperial’s wider Feedback and Learning: AI-Assisted & Reimagined (FLAIR)initiative, which explores how AI can be used to enhance feedback practices while preserving the relational, disciplinary and human-centred approaches to feedback. Rather than treating GenAI as a disruptive novelty, the event positioned it as a catalyst for rethinking feedback as a process – one that is collaborative, iterative and embedded within broader learning design.

“Our graduates will still need strong disciplinary expertise, but they must also learn to work with AI as a partner.” (Professor Alan Spivey, Interim Vice-Provost for Education and Student Experience) “When looking at how we develop feedback in a world of AI, we need to take our ÌìÃÀ´«Ã½ with us and work with them to get those solutions.” Professor Alan Spivey Interim Vice-Provost (Education and Student Experience)

Aims of the Conference: Rethinking Feedback in an AI Era

The central aim of the conference was not simply to ask what AI can do in feedback, but to consider what feedback means in an AI-enabled educational environment. Across the day, discussions explored questions such as:

• How is GenAI reshaping ÌìÃÀ´«Ã½’ engagement with feedback?
• What does ‘feedback’ mean for us in the age of AI?
• How can AI support, rather than replace, human judgement?

These questions reflect a broader institutional concern: ensuring that graduates are not only capable of using AI tools, but also able to think critically and independently within learning environments where AI tools are becoming more widely used. The conference emphasised that feedback remains central to this mission – but must evolve from a static output into a dynamic learning process.

Insights from the Keynote Panel Discussion: Feedback as Care, Process and Dialogue

“In the past, we used to have the knowledge constructions maybe in the form of human-human interaction. Now we have knowledge co-constructions with AI interactions. what does knowledge mean in this age of AI?” (Keynote panel discussion)

The keynote panel session provided a rich set of perspectives on how GenAI is already influencing feedback practices across higher education.

A major conceptual shift was the reframing of feedback as a process, rather than a set of comments delivered at the end of assessment. Within this view, feedback only becomes effective when ÌìÃÀ´«Ã½ actively engage with it by interpreting, questioning and applying it over time. Simply producing more feedback, whether human or AI-assisted, does not necessarily improve learning outcomes.

Instead, the focus is shifting toward: 

• feedback design rather than feedback volume
• student engagement rather than staff output
• learning processes rather than assessment endpoints

 These perspectives underpin much of the work within the FLAIR initiative, which seeks to integrate AI in ways that enhance feedback practices rather than automate isolated tasks.

Speakers emphasised that while AI tools can provide fast, structured and often highly readable feedback, they tend to lack the relational and contextual understanding that defines effective educational dialogue. It highlights a key takeaway: feedback is not only about information transfer, but also about recognition, trust and academic development.

How Students Are Already Using AI in Feedback Processes

A recurring theme across presentations was the reality that ÌìÃÀ´«Ã½ are already integrating AI into their learning workflows.

Evidence discussed during the event suggested that:

• a significant proportion of ÌìÃÀ´«Ã½ already use AI tools to interpret or refine feedback
• ÌìÃÀ´«Ã½ value AI for clarity, immediacy and privacy
• trust in AI-generated feedback remains uneven
• effective feedback relies heavily on human connection and accountability

This raises an important pedagogical challenge: ÌìÃÀ´«Ã½ are not waiting for institutional policy to catch up. They are already experimenting with AI as a feedback tool.

However, concerns were raised about shallow engagement, where ÌìÃÀ´«Ã½ may use AI to summarise or rephrase feedback rather than deeply engage with its meaning. This has been described as a risk of reduced metacognitive engagement, where reflection and planning are outsourced rather than developed.

As a result, developing feedback literacy, which is understood as the ability to interpret, evaluate and act on feedback (including AI-generated feedback), was identified as a key priority.

Risks, Ethics, and Institutional Responsibility

Alongside the opportunities, the conference also foregrounded several risks that must be addressed as AI becomes more embedded in feedback systems: 

• Hallucinations and inaccuracies in AI-generated outputs
• Bias and fairness concerns in automated systems
• Privacy and data protection issues
• Environmental costs associated with large-scale AI use
• Lack of clear institutional policy frameworks

These challenges reinforce the importance of transparency and governance in any AI-enabled feedback system. It also highlights that while AI can support educational practice, it cannot carry institutional responsibility, which remains firmly with educators and universities.

FLAIR and the Future of Feedback Practice

The FLAIR initiative at Imperial provides the institutional context for these discussions. Its aim is not to introduce AI into feedback for its own sake but to explore how AI can meaningfully enhance feedback design, feedback engagement and ÌìÃÀ´«Ã½’ capacity to use feedback in their learning.

This initiative creates space for open and critical conversations about the opportunities this presents, as well as the risks and questions that remain. While there is no single model for what human-AI collaboration in feedback practice should look like, FLAIR highlights the value of continued experimentation, reflection and shared learning with ÌìÃÀ´«Ã½.

Post-Event Resources and Continuing the Conversation

While the conference itself has concluded, the discussion it generated is ongoing. A range of post-event resources, including the keynote recording and supporting materials, are available for those who wish to explore the themes further.

The keynote panel video in particular offers a valuable opportunity to engage directly with the perspectives shared during the event and to hear how educators are thinking through the evolving relationship between AI and feedback in real time.

These resources provide an important extension of the conference, allowing the ideas developed during the day to continue informing practice across the sector.

Concluding Reflections

The Imperial FLAIR conference makes clear that GenAI is not simply changing how feedback can be delivered but also how feedback is understood.

Rather than replacing educators, AI is prompting a deeper reconsideration of what feedback is for: not just correction or evaluation, but learning, development and recognition.

The challenge ahead is not whether AI should be part of feedback practice, but how it can be integrated in ways that preserve the human relationships at the heart of education while enhancing the learning experience for ÌìÃÀ´«Ã½.