Dynamic Constraints Relaxation as a Theory of Insight

Abstract

Research on insight has accumulated empirical evidence on its cognitive processes. However, there is little agreement on what problem-solvers learn from their initial failures and at what point an insight actually takes place. To explore these issues, we first propose a general framework that involves three constraints, object-level, relational, and goal. The object-level and relational constraints represent people's natural preferences of how objects and relations in a given problem are represented. The goal constraint evaluates a degree of match of the current state to the goal, and leads problem-solvers to select specific combinations of the representations of objects and relations. In the processes of insight, these constraints operate simultaneously and are gradually relaxed by repeated impasses. Using a geometric puzzle problem, we empirically tested hypotheses derived from the framework. Experimental results revealed that the initial persistence in a wrong approach could be explained by the object-level and goal constraints, and that subjects could reach an insight by relaxing the object-level constraints as well as allowing easy operation of goal constraints.