Christoph Mathys on ‘Madness as a failure of active inference’

Madness as a failure of active inference

Abstract: Madness takes many forms, most of which can be explained as a failure of active inference. I will take overfitting, the core problem of all inference, as my point of departure and discuss its relation to psychosis. Neurosis, a harder problem, forces us take into account action, and active inference. I will give a brief introduction to my thoughts on this topic as they have developed since my last presentation at the NPSA group three years ago.

Bio: Christoph Mathys is Assistant Professor of Neuroscience at Scuola Internazionale Superiore di Studi Avanzati (SISSA) in Trieste, Italy. He has a PhD in information technology from ETH Zurich and a master’s degree in psychology and psychopathology from the University of Zurich. He is a candidate with the Swiss Psychoanalytical Society and has completed a four-year postgraduate training programme in psychoanalysis. During his graduate studies, he developed the hierarchical Gaussian filter (HGF), a generic hierarchical Bayesian model of inference in volatile environments. Based on this, he develops and maintain the HGF Toolbox, a Matlab-based free software package for the analysis of behavioural and neuroimaging experiments. His research focus is on the hierarchical message passing that supports inference in the brain, and on failures in this kind of predictive coding in relation to psychopathology.