Conati, C., Larkin, J., & VanLehn, K. (1997). A computer framework to support self-explanation. In R. du Boulay & R. Mizoguchi (Eds.), Proceedings of the Eighth World Conference of Artificial Intelligence in Education (pp. 279-286). Amsterdam: IOS Press.

We present a computational framework for improving learning from examples by supporting self-explanation - the process of clarifying and making more complete to oneself the solution of an example. Many studies indicate that self-explanation can improve problem solving performance, and that guiding self-explanation can extend these benefits. Our goal is developing and testing a computer tutor Ñ the SE (Self-Explanation) Coach Ñ that can elicit and guide correct and effective self explanation, and thus improve problem solving performance in university-level Newtonian physics, a particularly complex and psychologically challenging domain. The self-explanations elicited by the SE Coach address how each component in the example solution can be justified in terms of (a) the theory of the instructional domain, and (b) the goal accomplished in the plan underlying the example solution. The SE Coach provides the student with a Workbench that interactively presents examples and provides tools to construct self-explanations using the instructional domain theory. To guide self-explanation responsively, the SE Coach relies on a probabilistic student model, from which it assesses the student's understanding of an example. The student model consists of a Bayesian network that generates its predictions by integrating information on self-explanations performed in the Workbench with information on the student's general domain knowledge and on the structure of the current example. By examining this network, the SE Coach identifies deficits in the student's self-explanations and can provide guidance to remedy them.

For a PDF full article version, click here (224KB).