Abstract

The concepts of plan and planning and their applications have been evolving in the history of cognitive psychology and artificial intelligence. The classical assumption that plans govern actions has dominated the research in both domains for decades. However, the view is inadequate to account for complex and dynamic activities in real-world situations. While strongly refutes the controlling role of plans in actions, Suchman (1987) posits her provocative conjecture that plans serve only as prospective and retrospective representations of situated action, but do not determine the actual course concerned. Undoubtedly, this postulation poses a great challenge to the traditional cognitive theory and has its nontrivial implications to many scientific disciplines for which planning strategies are deemed essential. Computer programming is composed of a set of high-level cognitive skills of which debugging is particularly demanding. Based on the findings of some expert-novices studies, a group of researchers propose the plan-based theory. Its major claims are that possession of "programming plans" is the determining characteristic of programming expertise and such plans are language and problem independent. However, some recent empirical data have revealed the pitfalls of such a theory. It has been found that programming expertise should depend on a cluster of interactive factors of which programming strategy is the significant determinant. Specifically, the present review sets itself to examine the role of plan and planning in programming debugging and to propose an appropriate means for teaching this inherently situated skill. Cognitive apprenticeship grounded upon the theory of situated cognition appears to be a promising alternative. Tentative empirical designs for testing this idea and evaluating the veridicality of Suchman's theory have been suggested. While defying the existence of plans is prima facie absurd, the ultimate concern is to understand what kind of resources they are.