About the Project
Almost everything we know about how animals learn comes from one task at a time, in different animals. So a basic question is unanswered: are an individual's abilities separate skills, or expressions of a few underlying computations? If a rat learns quickly to accumulate evidence, is it also flexible when the rules change? Is it more social? These questions remain open because several abilities are rarely measured in the same individuals.
Here, the same rats will be trained across a battery of behaviours using a fully automated, high-throughput training platform: accumulating noisy sensory evidence toward a decision, switching between opposite rules on a cue, pure perceptual discrimination, and social interaction. Because training runs continuously and in parallel, each animal contributes hundreds of thousands of trials, enough to characterise one individual rather than an averaged group.
The analysis uses a new AI-based approach developed in the lab. Instead of fitting a model chosen in advance, large language models are placed inside an evolutionary search: candidate models are written as short programs, fitted to a single animal's choices, scored on how well they predict future behaviour, and mutated over thousands of iterations. What emerges is a mechanistic account as readable code, with parameters estimated per individual.
The student will then ask the central question: do the parameters that describe one animal in one task predict its behaviour in the others? Is there a small set of latent dimensions, a cognitive fingerprint, that travels with the individual across domains?
The student will gain a broad set of skills: automated behavioural hardware and experimental design, machine vision and pose tracking, reinforcement learning and recurrent network models, large-scale data analysis, and hands-on experience using large language models for scientific discovery. The project is co-supervised across experimental and computational neuroscience, and the student will be supported to publish and present internationally.
No previous experience with animals is required. The project would suit a quantitatively minded student from computer science, physics, engineering or psychology who wants to work on real behaviour, or an experimentalist who wants to build strong computational skills.
This 4 year PhD project is part of a competition funded by EASTBIO BBSRC Doctoral Training Partnership (DTP). Detailed guidance on the application process, and the EASTBIO Application and Reference Forms can be found at How to Apply | Biology.
Please send your completed EASTBIO Application Form along with a copy of your academic transcripts (as a pdf) to INCR.PostgraduateRecruitment@ed.ac.uk. You should also contact your referees and ask them to submit their references (on the EASTBIO reference form template) to INCR.PostgraduateRecruitment@ed.ac.uk by the application deadline of 16th November 2026.
Funding Notes
This opportunity is open to UK and international students and provides funding covering stipend and UK level tuition fees. The University of Edinburgh covers the difference between home and international fees meaning that the EASTBIO DTP offers fully-funded studentships to all appointees. There is a cap on the number of international students the DTP recruits. It is therefore important for us to know from the outset which fees status category applicants will fall under when applying to our University.
Please refer to UKRI website and Annex B of the UKRI Training Grant Terms and Conditions for full eligibility criteria.
Project supervisors
Dr Marino Pagan
Dr Marino Pagan's profile is coming soon.
Dr. Angus Chadwick
Career overview
Dr. Angus Chadwick is a Lecturer in Computational Neuroscience and AI at the School of Informatics, University of Edinburgh. They are affiliated with the Institute for Adaptive and Neural Computation.
Research interests
Dr. Chadwick's research focuses on computational neuroscience and artificial intelligence. They are particularly interested in the intersection of these fields, exploring how computational models can enhance the understanding of neural processes and contribute to advancements in AI.
Computational Neuroscience, Artificial Intelligence, Neural Computation, Adaptive Systems, Machine Learning, Cognitive Modelling, Brain-Inspired Computing