Giuseppe Giglia, Dimitri Ognibene
Perception is a complex, neural mechanism that requires organization and interpretation of input meaning and it has been a key topic in medicine, neuroscience and philosophy for centuries. Gestalt psychology proposed that the underlying mechanism is a constructive process that depends on both input of stimuli and the sensory-motor state of the agent. The Bayesian Brain hypothesis reframed it as probabilistic inference of previous beliefs, which are revised to accommodate new information. The Predictive Coding Theory proposes that this process is implemented through a top-down cascade of cortical predictions of lower level input and the concurrent propagation of a bottom-up prediction error aimed at revising higher level expectations. The ‘Active Inference’ theory explains both perception and action, generalising the prediction error minimisation process. In this focused-review we provide a historical overview of the topic and an intuitive approach to the new computational models.
