ABSTRACT
People who solve the Tower of Hanoi start out with a guided trial-and-error strategy and later acquire a recursive strategy, the generally most effective strategy. Protocol data shows that noticing and using sub-towers in problem-solving differentiates two subjects who acquired the recursive strategy from one who did not. A working Soar model explains Tower of Hanoi strategy-acquisition by first assuming the basic ability to notice and use subtowers, and then charting the process by which this new knowledge is integrated with existing knowledge to produce the recursive strategy. Of particular importance in the integration is learning to see nested subtowers and using simple spatial-manipulation reasoning to figure out how to move those subtowers. The model shows a good qualitative fit to the data, providing support for Soar as a unified theory of human cognition.
