Two Theories of Predictive Cognition

Abstract

Nowadays, representations are all the more often conceived as probability distri­butions. Although many authors continue to adhere to “conceptual” and “proposi­tional” idioms, meanwhile, there are cognitive theories that can explain the tran­scend of mental operations beyond the narrow framework of associations and reflexive learning to conceptual cognition. The article discusses two concepts based on the understanding of cognitions as probabilistic predictions: the theory of pre­dictive processing (PP) and the theory of functional systems (FS). The PP-theory relies heavily on the concepts of computation and representation. However, repre­sentations are considered here as sub-symbolic, and, accordingly, computations are understood as probabilistic (Bayesian) inference. The FS-theory believes that a cognitive system anticipates parameters of the expected result and strives to maintain a certain level of metabolism, as well as to minimize the degrees of freedom of the system. As it is shown, we don’t have sufficient reasons to con­sider both theories equivalent. They are located at different levels of the Marr’s scheme, use different mathematics, and, probably, are not equally falsifiable. Based on these considerations, it can be assumed that the explanations and pre­dictions of these two theories will differ.