Human Problem Solving
| Hi-index | 0.00 |
Production systems (PSs) can be written which are structurally adaptive in the sense of exhibiting different strategies, or mixtures of strategies, depending on the circumstances of the task. They consist essentially of the union of the individual PSs for each strategy. Such PSs begin to approximate the flexibility and adaptiveness of skilled human performance. An earlier analysis of children's seriation behavior is discussed in this light, and a PS for doing subtraction in a flexible manner is developed. It raises a number of issues concerning the user of PSs as models of cognitive processes: the role of explicit rule-ordering; the representation and retrieval of facts; and the possibility of eliminating the need for explicit "housekeeping" rules. The question of ordering is particularly interesting, since it appears that, on the one hand, explicit ordering is needed in order to specify fully a particular method, while on the other hand a more faithful model results it ordering is ignored so that certain aspects of the behavior are left unspecified.According to Newell and Simon [5], "production systems are the most homogeneous form of programming organization known". One consequence of this homogeneity is that a production system (PS) for solving a problem can be written in such a way that each individual rule represents an independent fragment of potential activity on the part of the problem solver and generates a meaningful component of the total problem solving process. Properly exploited, this independence enables one to write PSs which, instead of being committed to one particular strategy for tackling the problem, consist essentially of a mixture of different strategies and display an appropriateness and flexibility of behavior that begins to approximate the characteristics of skilled human performance. Such PSs are adaptive in the sense that the method they follow on a particular occasion is determined by the requirements of the individual task. I will refer to them as structurally adaptive PSs, in order to distinguish them from Waterman's "Adaptive Production Systems" [8]. Waterman's PSs adapt to a given problem by modifying themselves during the course of solving the problem, acquiring new rules that extend the original PS. The PSs described here do not do this. They remain fixed for the duration of any one task, and adapt to the problem by virtue of their inherent structure.In this paper, after illustrating these notions by a summary of an earlier study of children's seriation behavior (Section 1), I will present a structurally adaptive PS for subtraction (Section 2) and then discuss the issues it raises for the use of PS as psychological models (Section 3).