Modeling incubation and restructuring for creative problem solving in robots.

Creativity represents the pinnacle of higher-level cognition, but exactly how it is achieved remains poorly understood, especially when simultaneously facing the opposing challenge of intractable complexity. The aims of the current study were (a) to examine how the brain may achieve the dual goals of creativity and complexity reduction, and (b) to begin developing higher-level cognition and creativity in robots. We address these aims by (a) modeling an example of insight problem solving and comparing it to empirical data, and (b) testing the model on a robot platform. Unlike other models, we propose a single mechanism for both creative problem solving and complexity reduction. Focusing on creativity, the computational mechanism leads to insightful problem solving by restructuring an internal belief representation based on evidence collected during an incubation period. Because insightful problem solving has been examined closely in nonhuman primates, providing detailed quantitative datasets lacking in humans, we tested the model by comparing simulations to insightful problem solving by rhesus monkeys. Results show that the proposed model accounts for both the discontinuous three stage problem-solving patterns and the spontaneous generalization to novel cases observed with the monkeys. To test the model in a physical environment, we implemented it in a vision-equipped robot, and the model solved the same insight problem from camera percepts. Our model shows how the creative brain may address the dual challenges of complex environments–finding unprecedented opportunity hidden within potentially intractable complexity–and suggests that both challenges may be met by a single underlying computational mechanism. [ABSTRACT FROM AUTHOR]