Dynamic Constraint Relaxation Approach

In a past decade, more and more researchers have emphasized the emergent nature of cognition, as opposed to the traditional approach that assumes fixed representations and central control mechanism. Among many kinds of emergent cognition, insight has attacted many researchers’ attention (Sternberg & Davisdon, 1995). Insight has several mysterious properties. First of all, problems used in psychological experiments on insight is far from complex. Many subjects can understand the solution immediately after they are taught. However, it is awfully difficult to find it by themselves. Second, people stick to a wrong approach and repeated same errors. When solving a standard non-insight problem, people usually swith to a different strategy or search other paths after noticing failure. But, they do not do so in insight problem-solving. Third, they ignore useful information accidentally found or generated, and only later they realize its usefulness. Finally, insight appears to come to mind suddenly. These mysterious properties prevent the standard problemsolving framework from providing a coherent account. In order to deal with these problems, we have developed the dynamic constraint relaxation theory of insight (Hiraki & Suzuki, 1998; Suzuki & Hiraki, 1997). This theory consists of three kinds of constraints (objectlevel, relational, and goal), and a relaxation mechanism. The main idea is that impasses are formed by these constraints and that qualitative changes are caused probabilistically by the failure-driven incremental relaxation of these constraints. ∗The researches listed in this paper have done in collaboration with Kazuo Hiraki, Michiko Miyazaki, and Keiga Abe. This research was supported in part by Grant-in-Aid for Scientific Research (C)(No. 10610082) and by Grant-in-Aid for Specially Promoted Research(No. 14001003).