From AI-assisted supervisions to new approaches to assessing programming skills and creativity, six projects will investigate how universities can adapt teaching and learning for an age of AI.
The University of Cambridge has selected six projects for funding through ai@cam’s new AI for Teaching and Learning initiative. The projects will explore how AI can support university teaching, while helping students develop the knowledge, skills, and judgement they need in a world where AI is increasingly available.
The projects span architecture, economics, genomic medicine, law, physics, and psychology. They will test new approaches to teaching and assessment, examine how AI affects students’ learning, and develop practical resources that could be used more widely across the University.
Generative AI is already changing how students study, from summarising scientific papers and generating computer code to exploring literature and ideas. This creates opportunities for new forms of teaching and feedback, but also raises questions about how students develop expertise and learn to work effectively with AI.
The six projects approach these questions from different disciplines. Some will test how AI can help supervisors provide more individualised feedback or help students understand a topic. Others will investigate how assessment needs to change when AI can generate sophisticated outputs, or how teaching can help students retain responsibility for decisions made with AI assistance.
Professor Neil Lawrence, Chair of ai@cam, said: “AI is changing what students can do, but we still have a great deal to learn about what that means for education. Universities need to understand how these tools affect the development of knowledge, expertise and independent judgement, as well as where they can make teaching more effective.”
Jessica Montgomery, Director of ai@cam, added: “These projects give Cambridge educators an opportunity to test different approaches in practice and share what they learn across the University. The findings should help us make better-informed decisions about how AI is used in teaching and learning.”
The six funded projects
How AI affects creativity in architecture and design
Led by Dr Michal Gath-Morad, Department of Architecture
Generative AI tools can produce architectural layouts and visualisations in seconds. But does using them make students more creative, or encourage them to settle too quickly on familiar designs?
This project will investigate how AI affects creativity at different stages of architectural design. Students will complete comparable design tasks with and without AI assistance, covering both early-stage layouts and the rendering of developed designs. Independent experts will assess the originality, variety, and suitability of the resulting work.
The findings will help educators understand when AI supports creative thinking, when it may constrain it, and how students can learn to use these tools without losing control of their design choices.
Developing scientific literacy through an AI Socratic coach
Led by Matthew Weatherhead, Department of Psychology
AI tools can summarise scientific papers and explain complex findings, but relying on those summaries may mean students miss opportunities to develop their own analytical skills.
This project will develop an AI-based Socratic coach that helps psychology students engage critically with scientific literature. The tool will ask questions about research methods, evidence and interpretation, encouraging students to explain and refine their reasoning.
The team will compare students using the coach with those undertaking equivalent structured exercises without adaptive AI feedback. The aim is to establish whether AI-supported practice improves students’ ability to interpret and evaluate unfamiliar scientific research independently.
AI-augmented supervisions: supporting Cambridge’s small-group teaching
Led by Professor Will Handley and Toby Lovick, Institute of Astronomy
Cambridge’s supervision system gives students the opportunity to discuss problems and develop their understanding in small groups. This project will investigate whether AI can help supervisors make better use of that time.
The team will develop and pilot a Cambridge-hosted system that can record supervision discussions and whiteboards, help supervisors identify difficulties in submitted work, and make explanations available for students to revisit afterwards.
Working with around 20 supervisors across the physical sciences, the project will test whether AI-assisted preparation and feedback can improve the quality of teaching while preserving the personal interaction and independent reasoning at the heart of the supervision system. It will also examine practical questions around privacy, accuracy and student consent.
Seeing How We Think: mock law supervisions in an age of AI
Led by Dr Antoinette Nestor, Girton College
Generative AI can produce convincing legal arguments, including references to cases that do not exist or inaccurate accounts of legal decisions. For students still learning the law, these errors can be difficult to recognise.
This project will develop mock law supervisions designed to strengthen students’ ability to verify sources and defend their reasoning. Students will examine plausible but potentially unreliable AI-generated legal research, check claims against original authorities, and respond to questions that challenge their conclusions.
The project will produce a practical toolkit for supervisors. Its wider lessons could be useful across subjects in which students must learn to distinguish credible evidence from apparently authoritative answers.
Developing independent judgement in genomic medicine
Led by Timothy Hearn and the Cambridge Genomic Medicine Programme teaching team
AI is increasingly used to analyse genomic data, interpret evidence ,and support clinical communication. This creates new demands for training: students need to understand how to use AI effectively while retaining the ability to check its outputs and exercise their own professional judgement.
The project will redesign teaching activities in a course spanning bioinformatics, genomic variant interpretation, and genetic counselling. Students will use AI to support tasks such as developing analysis code, preparing clinical interpretations, and explaining genomic results, before demonstrating their understanding through independent exercises and discussion.
By comparing activities with and without AI support, the team will examine how these tools affect learning and identify approaches that help students develop the skills needed in genomic medicine.
If AI writes the code, what should we assess?
Led by Professor Peter Bossaerts, Faculty of Economics
When students can use AI to generate sophisticated computer programs, assessing the code they submit may reveal less about what they have actually learned.
This project will test a different approach in Cambridge’s Algorithmic Trading Robot Design course. Students will develop computer-based trading strategies and test them in a simulated financial market. Assessment will focus on how their trading robots behave under unfamiliar conditions, rather than simply examining the code.
AI-generated diagnostic feedback will help students identify weaknesses, revise their strategies and test them again. The project will investigate whether this iterative approach improves students’ understanding and offers a better way to assess programming-based work in an age of AI.
Learning what works
The AI for Teaching and Learning initiative supports Cambridge educators to experiment with new approaches to teaching and assessment, evaluate their effectiveness, and share the results with colleagues.
Over the coming six months, the funded teams will work with students and teaching staff to develop and test their approaches in existing courses and supervision settings. Their findings will contribute to a wider understanding of how AI can be used responsibly and effectively in university education.
The projects will share their findings, practical resources and lessons learned with the wider Cambridge community, helping inform future decisions about AI in teaching and learning.
Find out more: https://www.ai.cam.ac.uk/ai-in-the-university/