A Computational Model of How Learner Errors Arise from Weak Prior Knowledge

Matsuda, N., Lee, A., Cohen, W. W., & Koedinger, K. R. (2009). A Computational Model of How Learner Errors Arise from Weak Prior Knowledge. In N. Taatgen & H. van Rijn (Eds.), Proceedings of the Annual Conference of the Cognitive Science Society (pp. 1288-1293). Austin, TX: Cognitive Science Society.

How do differences in prior conceptual knowledge affect the nature and rate of learning? To answer this question, we built a computational model of learning, called SimStudent, and conducted a controlled simulation study to investigate how learning a complex skill changes when the system is given "weak" domain-general vs. "strong" domain-specific prior knowledge. We measured learning outcomes with the rate of learning, the accuracy of learned skills (test scores), and the accuracy in predicting patterns of real student errors. We found not only that the accuracy of learned skills decreases when weak prior knowledge is given, but also the learning rate significantly slows down. The accuracy of predicting student errors also increased significantly, and SimStudent with the weak prior knowledge made the same errors that real students commonly make. These modeling results help explain empirical results connecting prior knowledge and student learning (Booth & Koedinger, 2008).

PDF file