Research Project
Computational Models of Interoception in Psychiatric Disorders
Using Bayesian active inference frameworks, we model how individuals with depression, anxiety, eating, and substance use disorders fail to adaptively regulate interoceptive precision estimates. This transdiagnostic approach reveals shared computational mechanisms across psychiatric conditions.
Methods Used
- Active inference / hierarchical Gaussian filter
- Bayesian model comparison
- large transdiagnostic cohort (Tulsa 1000)
- fMRI
Key Findings
Transdiagnostic failure to adapt interoceptive precision confirmed across anxiety, depression, eating, and substance use disorders (PLoS Comput Biol 2020; replicated in 2024, Biological Psychology). An Annual Review of Clinical Psychology paper provides the theoretical framework.
Key publications
Ann Rev Clin Psych 2019 (doi:10.1146/annurev-clinpsy-050718-095617); PLoS Comput Biol 2020 (doi:10.1371/journal.pcbi.1008484); Biol Psychol 2024 (doi:10.1016/j.biopsycho.2024.108825)
Funding source(s)
2P20GM121312 (NeuroMAP Center Grant, NIGMS)
Collaborators
- Justin Feinstein (LIBR);
- Martin Paulus (LIBR);
- Murray Stein (UCSD)