ERIC BRUNET-GOUET

CURRENT POSITION

Psychiatrist at the Adult Psychiatry Department of Versailles Hospital

MD, PhD, HDR

X: https://x.com/ericbrunetgouet

Orcid: https://orcid.org/0000-0002-3784-7817

Research Gate: https://www.researchgate.net/profile/Eric-Brunet-Gouet

email: ebrunet (at the domain name) ght78sud.fr

Area of Research

Schizophrenia, Cognition, Neurosciences

AI technologies in mental heath

CURRENT WORK

My main research interest concern severe and persistent mental conditions associated with psychological disability. I have been working on social cognition evaluation (theory of mind) in schizophrenia, developping experimental paradigms for behavioral and neuroimaging studies. The conclusions are now published. We are currently working on a hybrid study combining clinical research and the implementation of a psychosocial rehabilitation intervention in several adult psychiatric outpatient centers in the west part of Paris (PASSVers2).

In the past years we have been working with Pr. Paul Roux on larger populations of patients assessed by the FondaMental Center of Expertise. This work raised the interest for data processing methods allowing us to manage large and unstructured datasets. Deep learning algorithms, natural language processing (for instance, deep-attentional or transformers models or lstm models) count among the most innovative and promising ways to investigate mental health questions. We are currently working with a BERT-derived model as shown in this article :

In another work, we challenged the recently proposed ChatGPT model with theory-of-mind tasks in order to see whether it had some mentalizing skills or not. Here we report some contrasted results, the chatbot demonstrating impressive answer generations skills. However, it seems sometimes to miss completely the use of an intentional stance. We discuss some improvements in the training procedure in order to favor mentalization logic.

Investigating large databases with complex data raises several questions about data encoding and the appropriate way to manage this complexity. We recently started with Nathan Vidal a work about prescription data in mental health and more precisely the pharmacological effects of psychotropes. It is well known that anticholinergic properties of antipsychotics and antidepressants are negative for cognitive functioning. Nathan will carefully consider this question and provide some new ways to encode treatments and their differents properties in order to process large datasets from hospitals or outpatients facilities with these predictors.

COLLABORATIONS

RECENT Publications

Book Chapters

EXTERNAL LINKS