Enhancing PIXIE's Tutoring Capabilities
Abstract
PIXIE is an Intelligent Tutoring System shell that attempts to diagnose and remediate student errors in a particular domain (Sleeman, 1987). This system has been implemented as three separate subsystems. The first, the offline phase, generates models that incorporate typical bugs, or errors, in the domain to be tutored. These bugs have been collected through paper-and-pencil tests and interviews with students. The second, the online phase, uses the models previously generated in the offline phase to diagnose and remediate a particular student's errors. The overall structure of the PIXIE system is shown in Figure 1. The separation of the system into three system into three phases allows for a quick response time during the actual tutoring session; however, it also means that PIXIE is only able to detect previously encountered bugs, or mal-rules. Implementing a system capable of diagnosing bugs that have not been encountered before is a research topic currently being pursued. Presently, unanticipated answers can be processed during the third phase, post-interaction analysis, and, if consistent, added to the domain knowledge base. Keywords: Artificial intelligence, Expert systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1988
- Accession Number
- ADA199020
Entities
People
- D. Sleeman
- J. L. Moore
- Judith Orasanu
Organizations
- Stanford University