Charting trajectories of human thought using large language models
Status: This talk is in preparation - details may change
Status: This talk has been cancelled
For our next talk, in the Digital Phenotyping seminar series, we will hear from Matthew Nour, NIHR Clinical Lecturer in Psychiatry, Senior Research Fellow and Honorary Consultant Psychiatrist at the Department of Psychiatry, on Wednesday 22 October, 2:00 pm – 3:00 pm, at the Big Data Institute (BDI).

Title: Charting trajectories of human thought using large language models

Date: Wednesday 22 October 2025
Time: 2:00 pm – 3:00 pm
Venue: BDI/OxPop, Seminar Room 0; followed by refreshments in the atrium

Abstract:
Understanding how humans organize and sample conceptual knowledge is a central challenge in cognitive science. Language has long been recognized as providing our richest behavioural window onto thought, yet a reliable mapping of words to mental concepts has proven formidable. Artificial Intelligence large language models (LLMs) constitute highly sophisticated models of human conceptual organization, offering unprecedented opportunities to revisit the challenge of decoding thought content from words. We introduce VECTOR, a computational framework that transforms language model representations into cognitively-meaningful representational spaces. Applying VECTOR to narrative samples generated by 1,100 participants, we cast individual narratives as geometric trajectories through conceptual space, revealing how thoughts flow from one idea to the next. Crucially, these narrative trajectories have interpretable geometric properties, predicting momentary language behaviour (response times) as well as individual differences in real-world communication patterns. By revealing the hidden geometry of thought, VECTOR enables quantitative investigation of uniquely human mental processes, opening new avenues for understanding how humans dynamically organize and navigate knowledge in naturalistic settings.

Short bio
Matthew studied Medicine, Neuroscience and Philosophy at the University of Oxford. He subsequently combined postgraduate medical training with clinical neuroscience research, first as an Academic Foundation Doctor (Oxford), and later as an MRC Clinical Research Fellow (MRC London Institute of Medical Sciences & Imperial College London) and an NIHR Academic Clinical Fellow in Psychiatry (Institute Psychiatry Psychology and Neuroscience, King’s College London).

In 2018 Matthew was awarded a Wellcome Trust Fellowship to complete a PhD in Cognitive and Computational Neuroscience at UCL under the supervision of Professor Ray Dolan FRS (UCL) and Professor Zeb Kurth-Nelson (DeepMind).

Matthew re-joined the University of Oxford in 2022 as an NIHR Clinical Lecturer in Psychiatry, and is now a Senior Research Fellow and Honorary Consultant Psychiatrist at the Department of Psychiatry. His current interests span cognitive neuroscience, AI in psychiatry (“computational psychiatry”), and psychopharmacology.

Hybrid Option:
Please note that these meetings are closed meetings and only open to members of the University of Oxford. Please respect our speakers and do not share the link with anyone outside of the University. The purpose of these seminars is to foster more communication among employees throughout the University, so we strongly advise in-person attendance whenever feasible.

Microsoft Teams meeting
Meeting ID: 366 709 908 720 5
Passcode: G7jp3Aj7
Date: 22 October 2025, 14:00
Venue: Big Data Institute, Old Road Campus OX3 7LF
Venue Details: Seminar room 0
Speaker: Dr Matthew Nour (University of Oxford)
Organising department: Big Data Institute (NDPH)
Organiser: Sumeeta Maheshwari (University of Oxford)
Organiser contact email address: sumeeta.maheshwari@ndph.ox.ac.uk
Hosts: Dr Aiden Doherty (University of Oxford), Dr Anya Topiwala (University of Oxford)
Part of: Digital Phenotying
Booking required?: Not required
Booking email: sumeeta.maheshwari@ndph.ox.ac.uk
Audience: Members of the University only
Editor: Sumeeta Maheshwari