Heuristic reasoning about uncertainty: an artificial intelligence approach
Heuristic reasoning about uncertainty: an artificial intelligence approach
Using planning techniques in intelligent tutoring systems
International Journal of Man-Machine Studies
Intention-based diagnosis of errors in novice programs
Intention-based diagnosis of errors in novice programs
Knowledge-based tutoring: the GUIDON program
Knowledge-based tutoring: the GUIDON program
Artificial intelligence and tutoring systems: computational and cognitive approaches to the communication of knowledge
Student model revision: evolution and revolution
Proceedings of the eighth biennial conference of the Canadian Society for Computational Studies of Intelligence on CSCSI-90
AI Magazine
Context dependent planning in a machine tutor (artificial intelligence, teaching systems, meno-tutor)
Accretion Representation for Scrutable Student Modeling
ITS '00 Proceedings of the 5th International Conference on Intelligent Tutoring Systems
Hi-index | 0.00 |
This paper describes an approach to student modeling for intelligent tutoring systems based on an explicit representation of the tutor's beliefs about the student and the arguments for and against those beliefs (called endorsements). A lexicographic comparison of arguments, sorted according to evidence reliability, provides a principled means of determining those beliefs that are considered true, false, or uncertain. Each of these beliefs is ultimately justified by underlying assessment data. The endorsement-based approach to student modeling is particularly appropriate for tutors controlled by instructional planners. These tutors place greater demands on a student model than opportunistic tutors. Numeric calculi approaches are less well-suited because it is difficult to correctly assign numbers for evidence reliability and rule plausibility. It may also be difficult to interpret final results and provide suitable combining functions. When numeric measures of uncertainty are used, arbitrary numeric thresholds are often required for planning decisions. Such an approach is inappropriate when robust context-sensitive planning decisions must be made. Instead, the ability to examine beliefs and justifications is required. This paper presents a TMS-based implementation of the endorsement-based approach to student modeling, discusses the advantages of this approach for planner-controlled tutors, and compares this approach to alternatives.