dayuso@BBN.COM ("Damaris M. Ayuso") (06/20/89)
BBN STC Science Development Program AI Seminar Series Lecture KNOWLEDGE REPRESENTATION AND EXPLANATION IN AN INTELLIGENT TUTORING SYSTEM FOR PROGRAMMING Brian J. Reiser Princeton University bjr@confidence.Princeton.edu BBN STC, 2nd floor large conference room 10 Moulton St, Cambridge MA, 02138 Friday June 23rd, 10:30 AM I will describe an intelligent tutoring system for programming called GIL (Graphical Instruction in LISP) that constructs explanations directly from the content of its problem solving knowledge. GIL provides feedback by comparing a student's solution to its problem solving model. GIL's problem solving rules not only encode what step to take in each problem situation, but also can reason about why each step is effective. Explanations are constructed dynamically rather than being prepared in advance for each situation in which feedback is required. The tutor is embedded in a graphical programming environment so that students work in a medium that more closely corresponds to their planning operations. GIL produces reasonable explanations for a wide variety of errors concerning approximately 200 rules and high-level plans used in an introductory programming lesson. I will describe studies of students learning to program using GIL and working with human tutors, and consider: (1) how GIL's graphical representation facilitates students' reasoning, (2) how GIL's explanations enables students to learn more effectively from their errors, and (3) the ways in which the pedagogical strategies and effectiveness of human tutors are modeled in GIL.