Date of Award
Doctor of Philosophy (PhD)
Carnegie Mellon University
College of Humanities & Social Sciences
John R. Anderson, Ph.D.
Jill H. Larkin, Ph.D.
Herbert A. Simon, Ph.D.
cognitive, psychology, cognitive modeling, ACT-R, examples, memory, knowledge
This dissertation examines the way people acquire procedures from examples, and provides a computational model of the results. In four experiments, people learned an analog of algebra. For each experiment, the initial know ledge that people had of the task was varied. In two experiments (Experiments 1 and 3), the syntactic know ledge that people had concerning the task w as manipulated. The knowledge of syntax that participants had, particularly the ability to correctly parse the character string, was found to be a major determiner in the way participants acquired the rules. Experiment 2 explicitly manipulated participant's awareness as to how the task was related to their prior knowledge of algebra, with the finding that another major determiner of how the participants learned the task resting on how much of the task they can map to algebra. All three of these experiments examined the rule generalization behavior of the participants, with a fourth experiment specifically designed to examine this issue. The less syntactic and other declarative knowledge that participants had, the less general their rules. These findings, that people can learn from examples but that this learning is tempered by their additional declarative knowledge, are captured by an ACT-R model (Anderson, 1993).
Blessing, S. B. (1996). The use of prior knowledge in learning from examples (Doctoral dissertation). Retrieved from http://digitalcommons.imsa.edu/alumni_dissertations/11