A computational model of the role of prior knowledge in learning to operate an interactive device
TD Daniel Freudenthal · TU/e Research Portal · 2000
This paper describes a computational model that simulates the behaviour of subjects in an experiment where they learn to control an interactive device. The model was specified in the ACT-R cognitive architecture. The amount of foreknowledge that the subjects had was manipulated, and the resulting differences in behaviour were assessed. The model employs a combination of instance learning and difference reduction strategies. Subjects without foreknowledge start out exploring the device in a random fashion, which is replaced by more informed strategies as the number of successful instances increases. For subjects with foreknowledge, strategies are more informed from the outset. The model provides a reasonable fit to the number of actions that both groups of subjects require for later rounds, but is less successful in modelling the behaviour of very early attempts.