A General Systems Analysis of Goal-Directed Heuristic Computation

Abstract

The paper delineates dynamic structures in heuristic computation using a general systems approach. A design for goal-directed heuristics is put forward. Distinctions begin with a discussion of the involvement of goals in computation and their differing roles in algorithmic and non-algorithmic heuristics. AI problem solving is used as a paradigm area for justifying a non-algorithmic approach, in particular a goal-directed approach using current rather than a priori control. The main distinction is between goal-directed heuristics and algorithms according to procedural criteria as well as on the guarantee of success. A general systems model of goal-directed heuristics is developed with self-alteration being a key design principle.