Intelligent Tutoring for Programming Tasks: Using Plan Analysis to Generate Better Hints.

摘要

This project has dealt with tutorial systems for computer programming languages, particularly, systems in which a student is trying to write a computer program and can, upon request, receive hints about errors in his program and ways he can correct these errors. This research had two phases: an experimental investigation of the interaction between a student and a hint-giving tutor, and the construction of a tutorial system that identifies the plan underlying a student's program and gives hints that address errors in this plan. In the experimental work, an existing programming language tutor (BIP) was modified to allow students to request and receive hints from a human tutor. These experiments suggested that: (1) Programs are typically written in two stages: layout of the basic design of the program and (2) correction of the statements that instantiate this design. Systems that rely on 'canned' hints do not provide adequate assistance for the complex problems faced by students. Although BIP was able to give reasonable help on the syntactic form of language statements, it has no facilities for identifying and offering advice on problems with a program's design; in these experiments, students recognized this shortcoming and relied upon assistance from the human tutors.