Maximizing entropy of the Predictive Processing framework
The Predictive Processing (PP) framework offers a unifying view on the existence and working of all living systems. The core premise of PP states that as long as agents minimize prediction error, and consequently entropy, they are successful. Current developments and advances in PP indicate that the interaction between agents and their environments is an important component of entropy minimization. In this paper, we explore by means of computer simulations, the interaction between PP-agents and their environments under different conditions. We argue the need to redefine the notion of success in PP in terms of entropy, behavioral and cognitive success, as we show that the environmental conditions that lead to entropy success, are different from conditions that lead to behavioral or cognitive success. Furthermore, we show that being equipped in and applying the mechanisms to minimize prediction error, do not in practice guarantee that the agents will be successful in any sense (entropy, cognitive or behavioral).