Der problemlöser robbimat

Abstract

We first show how to formalize environments in which situations can be changed only by action, and in which a situation should be brought about that is acceptable. This leads us to the concept of a problem constellation that consists of an action system, a reaction system, a judgement system, and an initial situation. We then present a cybernetical system which we call Robbimat. It is capable of testing problem constellations, of modifying them according to the results of the tests, and of judging the final constellation. We designed Robbimat in order to have favourable problem constellations automatically generated. As an example we discuss a simple game. It constitutes a win-loss-environment the structure of which can be deduced from the payoff function. We show how Robbimat analyses such an environment, and utilizing its structure classifies the initial situation.