ABSTRACT
In this chapter I apply the conceptual coordination perspective on memory and learning to models of reasoning, or symbolic information processing. In considering models of remembering in Soar (section 2.2.2), MOPs (section 11.2), and EPAM (section 11.3), I have already show that knowledge is frequently represented hierarchically in descriptive cognitive models. Correspondingly, problem-solving architectures, which model human reasoning, contain one or more hierarchies of problem types and solution methods. These hierarchies provide a means of focusing attention and ordering reasoning steps. For example, the hierarchy of infectious diseases in a medical disease taxonomy would constitute a problem space for focusing on particular diagnoses. The contained subtree of bacterial infections constitutes a more specific problem space. Similarly, diagnostic methods, such as procedures for gathering information, forming hypotheses, and discriminating hypotheses, may be organized into hierarchies of problem-solving operators. This chapter concerns how such hierarchies are structured and related in descriptive cognitive models and what, if anything, this reveals about conceptual coordination at the neural level.1
Extending the perspective of studying Caitlin in the pool (chapter 5) and slips (chapter 6), my underlying question is: How can people conceptualize and perform procedures before they have a language for rationalizing what they do? We find that the descriptions of patterns in problem-solving models often go
