Dynamics, filtering and estimation in perceptual and cognitive agents
It will be discussed some broad questions about dynamical aspects of cognition, to understand what makes a system of processes a cognitive one. It will be addressed the roles of information, context and structural features of the underlying processes, looking for some characteristics for a suitable mathematical structure underpinning cognition. Three kinds of agents will be compared concerning the information-theoretic aspects of action control: the reactive, the perceptive and the cognitive agent. Then I’ll pave the way for comparing the essential features that distinguishes perceptual and cognitive processes, which lead to the geometric characterization of each one under a view based on Klein’s Erlangen Program, in the search for answers to the evolutionary question of how cognition may have emerged. It will be discussed the geometrical and topological aspects of the dynamics of a cognitive agent, which distinguishes it from merely reactive and perceptive agents. What singles out the cognitive agent is the use of context-invariant information that breeds knowledge. In this approach, the perceptual system is formed by a chain of filtering and estimation processes dealing with contextual information, beginning at the sensorial system and ending at the motor system. The perceptual chain is constrained not only at both extremities, but its transformations are geometrically connected with “holonomic” constraint propagation. On the other hand, the cognitive system comprises self-constrained filtering/estimation processes that operate with learned transformations, so that the geometry of cognition is “non-holonomic”. Referência: Kogler Jr, J.E. and Santos, P.E. (2017) – Information, context and structure in cognition. In Adams, F. et al, Cognitive Science: Recent Advances and Recurring Problems, pp.177-194, Vernon Press, ISBN:978-1-62273- 100-8, Delaware, 2017.