The objective of this research is to explore a few of the key cognitive science, educational, and software engineering issues surrounding the development of effective, computer-based, instructional systems. This exploration will take place in the context of building and evaluating a specific, computer-based, instructional system-DEBUG-ASSISTANT. It will provide high-school students with instant, on-line feedback on the non- syntactic bugs in their Pascal programs. In order to build a system capable of coping with (1) a semester's worth of assignments, and (2) the significant variability found in students' programs, a key question in cognitive science will be explored: can bugs and variability be predicted for a domain as complex as programming? In classroom tests of the DEBUG-ASSISTANT the educational utility of providing students with this type of on-line assistance will be examined. Finally, issues of interface design for educational software and software engineering practices as they relate to building modifiable, expert systems will be explored. While preliminary research suggests that bugs and variability can be predicted, and that providing feedback is educationally beneficial, the use of the DEBUG-ASSISTANT will enable an investigation of these issues in a more systematic and rigorous fashion. This effort will be conducted in cooperation with administrators, teachers and students of the New Haven School System.
