Computational Psychiatry

Theory-driven computational psychiatry involves the use of methods of neural reinforcement learning to understand dysfunctional emotion and decision-making. The hope is to generate new nosologies for disease, along with new and precise ways of interrogating problems that patients might have, the therapies that might have a higher chance of working, and for making individually accurate prognoses. It is also hoped that we will be able to take advantage of radically new sources of high frequency data, such as that available from the ever more ubiquitous smartphones. Computational psychiatry is a nascent area and, as it starts to come of age, is grappling with critical questions such as test-retest reliability.

The lab has worked in the field almost since its inception. Work in Tübingen will be helped by the fact that the Department of Psychiatry, in a bid led by Prof. Andreas Fallgatter, has become one of the six centres in the country associated with the German Centre for Psychiatric Research (DZPG), with an application that included computational psychiatry as one of its three foci. The funding for the DZPG is arriving more slowly than originally hoped; however, now that Tobias Hauser has started, I hope that stronger connections will be forged.

We have completed a number of projects in this domain, including a new formal treatment of risk sensitivity in sequential domains (Gagne & Dayan, 2022), which we are currently testing in an online experiment, and we are extending it to the analysis of the famous balloon adaptive risk task (BART) and to the substantial inter-individual differences in neophobia and neophilia in a population of mice studied by Mitsuko Watabe-Uchida (Akiti et al., 2022), and most recently to large language models.

We operationalized the attribution-self representation cycle theory in a simple on-line task, and studied the resulting complex dynamics (Zamfir & Dayan, 2022), and we are currently writing a review of such cycles from the perspective of metacognition. We have also continued to formalize aspects of theory of mind to provide theoretically-motivated ideas for examining social components of psychiatric disorders, including paranoia (Barnby, Dayan & Bell, 2023a; Alon et al., 2023) and personality disorders (Barnby et al., 2022). With Andreea Diaconescu (Univ. Toronto), we are looking at Pavlovian ‘misbehaviour’ (i.e., untoward non-contingent infuences) in the context of suicidality.