Adams, Rick A;
(2018)
Bayesian Inference, Predictive Coding, and Computational Models of Psychosis.
In: Anticevic, Alan and Murray, John D, (eds.)
Computational Psychiatry: Mathematical Modeling of Mental Illness.
(pp. 175-195).
Elsevier: Amsterdam, Netherlands.
![[thumbnail of Adams_Psychosis. and delusion formation chapter.pdf]](https://discovery.ucl.ac.uk/style/images/fileicons/text.png) |
Text
Adams_Psychosis. and delusion formation chapter.pdf Access restricted to UCL open access staff Download (315kB) |
Abstract
The brain is thought to be a hierarchical Bayesian model of its body and its environment that performs inference on the causes of its sensations using predictive coding. In such a model, the accurate encoding of precision (inverse variance) of both prior beliefs and sensory data is essential. If the precision of prior beliefs is reduced, then inference will be biased toward sensory data, e.g., chance events will cause unwarranted updates to higher-level beliefs. In schizophrenia both a hierarchical imbalance in synaptic gain and a loss of inhibitory function (i.e., increased “excitatory/inhibitory balance”) bias inference in this way. This change may underlie the increased learning rate seen in belief updating paradigms in schizophrenia, although this may not be the only contributor to the “jumping to conclusions” bias—a proposed factor in the formation of delusional beliefs. Potential factors underlying the maintenance of delusional beliefs are also discussed, including reversal learning problems, and attempts to model these processes computationally.
Archive Staff Only
 |
View Item |
