John O'Doherty
California Institute of Technology, CA·Neuroscience
Why do people differ so much in what they like and in how they learn from experience? And can those differences be measured precisely enough to say something useful about mental health?
- Fields
Neuroscience
Psychology
Medicine & Clinical Research
- Topics
computational psychiatry
decision-making
reinforcement learning
- Grants
- 2 known grants, through Jan 2030
What they work on
- Separate occasion-related noise, meaningful state changes and stable trait differences in computational parameters by administering a computational assessment battery repeatedly over one year online and fitting hierarchical Bayesian models.
- Test whether trait-like variance components predict psychiatric diagnostic category while state-related variance tracks current symptom severity, in a diverse Southern California cohort with depression, anxiety and OCD.
- Track computational parameter estimates alongside changes in symptoms over time in a subset of patients to see whether shifts in learning and decision parameters accompany clinical change.
- Test whether single neurons in the human amygdala and lateral orbitofrontal cortex encode individual features of foods, art and clothing, using intracranial electrode recordings in epilepsy surgery patients making liking judgments.
- Determine whether feature signals are integrated into overall value in ventromedial prefrontal cortex, and whether a change in goal context alters feature responses themselves or instead re-weights their connections to value-coding neurons.
More about the research
Both projects treat thinking as a set of specific mental computations that can be written as equations and estimated from behavior or read out from brain activity. One asks how reliably such computational measures can be extracted from an individual and whether they track psychiatric diagnoses and symptoms over time. The other records from single neurons in the human brain to see how a preference is actually assembled.
Why it matters
Reliable computational measures of how a person learns and decides could eventually give clinicians objective markers to complement interview-based psychiatric diagnosis and to follow whether symptoms of depression, anxiety or obsessive-compulsive disorder are improving. Recording directly from human neurons as people judge what they like may explain where value signals come from in the first place, informing understanding of everyday decision-making and of conditions in which preferences and choices go awry.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Leveraging computationally derived measures of individual differences in learning and decision-making to predict psychiatric diagnosis, symptoms and changes in symptom severity across time
NIH · Apr 2025 – Jan 2030
Neuronal substrates underlying the construction of value in humans
NSF · Oct 2023 – Sep 2027
Closest labs
Carol Mathews
University of Florida, FL·Psychology
Test whether a dimension from hypervigilance to perceptual avoidance, measured with visual evoked potentials and frequency-tagged steady-state EEG potentials, distinguishes obsessive-compulsive and anxiety diagnoses and predicts severity.
Damian Stanley
Adelphi University, NY·Psychology
Identify social computational phenotypes associated with mental health spectra in an exploratory online sample of 1,000 adults, combining computational models of social and nonsocial learning tasks with HiTOP symptom assessment.
Peter Rudebeck
Icahn School of Medicine at Mount Sinai, NY·Neuroscience
Determine how bottom-up influences on learning change single-neuron activity and communication between the ventrolateral prefrontal cortex and basolateral amygdala, by recording neurons in macaques performing a probabilistic stimulus-reward task.
The school
California Institute of Technology: 18 labs on PI Pages, $445M a year on research, 200 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.