Alongside our teaching, we study why the Kitchen model works and use the kitchen as a place for new research on food.
Computational gastronomy for sustainable protein recipe innovation

Sustainable proteins make sense nutritionally and environmentally, but they often struggle to become part of what people actually cook. The barrier is largely cultural. Current computational gastronomy tools learn from which ingredients appear together in existing recipes, and novel proteins have almost no recipe history.
With the Bezos Centre for Sustainable Protein, the Dyson School of Design Engineering and the Department of Computing, we are building computational models that go beyond recipe co-occurrence. They combine flavour chemistry, sensory measurement, cooking behaviour and nutritional composition to predict where new ingredients can fit within a given cuisine. The goal is to support new dishes that feel as though they always existed, not just substitutions in familiar ones. We plan to release the outputs open access, including models, data and a tool chefs and culinary students can use.
This work forms the doctoral research of Cyrus Howbrook, a PhD student at the Department of Chemistry, co-supervised across Chemistry, Design Engineering and Computing.
The project is funded by the
Educational research
Every Chemical Kitchen course is evaluated, and what we learn goes back into how the courses are designed. Our published studies cover chemistry, medicine and biomedical science. In the Chemical Kitchen, students who started with low confidence in their lab skills gained the most, and the gap between students closed by the end of the course. The Biomedical Kitchen showed a similar effect on self-efficacy. Medical Kitchen students reported better procedural skills, such as suturing, and a new awareness of non-technical skills.
So far, each study has looked at one course and was carried out by the team that designed it. Our next step is to explain why the approach works, for whom, and under what conditions. We are looking at five possible mechanisms, drawn from research fields that rarely talk to each other:
- Productive discomfort: the deliberate unfamiliarity of working in a kitchen.
- Enactive mastery: succeeding at a real task, which builds confidence.
- Reduced cognitive load: with the subject content removed, students have more attention for practical skills.
- Identity safety: a setting where learners can try out their professional selves without much at stake.
- Reflective mediation: structured reflection that turns experience into insight they can use elsewhere.
We expect these mechanisms to act together rather than alone. We are also interested in a wider question: can a setting that looks nothing like the profession teach as well as, or better than, one built to imitate it? Our courses vary in discipline, stage of study and type of learner, so comparing them lets us separate the essential features of the model from the incidental ones. We will also test what carries over when the approach is adopted by other disciplines and institutions, including in vocational education.