Probabilistic inference in psychosis and autism

Abstract

Within the predictive coding framework the brain is defined as an inference machine that continuously tries to predict its sensory inputs on the basis of beliefs about the world and updates those beliefs in the presence of contradictory sensory data (i.e. prediciton errors; Friston, 2005). Neurobiologically, the weighting and further processing of those prediction errors is thought to be influenced by the gain of neuronal error units (Friston, 2010). When explaining the aberrant cognitive processes in patients with psychosis and autism, models based on this account have generated contradictory predictions. One main question is if the patients’ beliefs are too imprecise, too precise, or if the weighting of prediction errors is aberrant. In our study we are trying to test these hypotheses directly, using two different tasks that measure the precision of the prior belief and the weighting of the prediction error.