Zhao Cheng, Laura Ray, Ha T. Nguyen, and Jerald D. Kralik
Insight problem solving, Reinforcement learning, Q-tree algorithm
The underlying mechanisms giving rise to insightful problem solving are largely unknown, although models have been proposed that suggest phases of incubation, restructuring, and insight. Here, we propose a single computational mechanism that models the dynamics of structuring percepts to create an internal belief representation and subsequent restructuring of that belief representation, in which evidence for restructuring a representation is accumulated through trial-and-error attempts to solve the problem. Restructuring gives rise to a macroscopically observable incubation period followed by sudden emergence of a solution. The computational mechanism is evaluated through modeling of the nine-dot problem, a classic insight problem. Evidence supporting the notion of partial reward, a key concept in the computational mechanism, is observed in the progression to a solution of subjects engaged in the nine dot problem. The results of the model show that complex, high-level cognitive processes may emerge from simple underlying computational mechanisms.
Important Links:
Go Back
