The study of learning and problem solving using artificial devices: Synthetic Epistemology

The multi disciplinary study of mind, brain and behavior has reached a point where we seem to be confronted with the dilemma of studying either the computational mind or the dynamical brain. The cognitivist study of the computational mind has shown to be very effective in describing many elements of high level cognition. In the mean time, however, its most explicit expression, classical AI, is facing a number of fundamental problems. This has created a situation where arguments are raised in favor of abandoning the computational view. In this paper this dilemma is described. It is demonstrated that only one problem is hidden behind the multiple challenges facing the traditional computational program; the problem of a prioris, it is shown that the central component in a solution of this problem focuses on the nature of the knowledge ascribed or implemented by cognitive systems. A research program is defined, synthetic epistemology, which aims at addressing this question. As an example of this approach the modeling series of Distributed Adaptive Control is described. It is shown that this modeling series, which reflects elements of the behavioral paradigms of classical and operant conditioning and has been shown to be consistent with aspects of the correlated neural substrate, provides an alternative unifying perspective to the dilemma facing the study of mind, brain and behavior.