Control for Intelligent Tutoring Systems: A Blackboard-based Dynamic Instructional Planner

Article type: Research Article

Authors: Murray, William R.

Affiliations: Artificial Intelligence Center, FMC Corporation 1205 Coleman Avenue, Box 580 Santa Clara, CA. 95052

Abstract: An intelligent tutoring system that delivers effective instruction must select pedagogical actions appropriate to its tutorial situation. The approach taken in this research is to view this control problem as a dynamic planning problem. Dynamic instructional planning is the ability to generate, monitor, and revise instructional plans during the course of instruction. Planning and execution of instructional actions must be interleaved because the tutor operates in a changing environment with incomplete information. An appropriate architecture is the BB1 Blackboard Architecture, which supports the building of knowledge-based planners that represent and reason about their own actions. Dynamic instructional planning uses these capabilities to apply pedagogical knowledge to reason about instructional actions that a tutor can perform. We have built the Blackboard Instructional Planner in BB1 to teach troubleshooting a complex physical device by first imparting a mental model of the device and its operation. The planner generates instructional plans from skeletal plans, executes them and monitors their effectiveness. Instructional plans are modified and particular instructional actions selected in response to changes in the student model, changes in resources available, requests and questions of the student, and to properties of the subject matter currently being presented.

DOI: 10.3233/AIC-1989-2201

Journal: AI Communications, vol. 2, no. 2, pp. 41-57, 1989