AAAI Conference 1986 Conference Paper
An Analysis of Tutorial Reasoning about Programming Bugs
- David Littman
New Haven, CT 06620 what knowledge to teach a student who makes a bug nor about how to teach the knowledge. In a sense, all the tutorial knowledge possessed by such systems is ‘ compiled”. The three step model may be appropriate for tutoring students when their bugs do not reflect deep misunderstandings, or when one bug always should get the same intervention. However, it seems unlikely to be effective in domains such as computer programming where students’ bugs are often related and may reflect deep misconceptions about how to solve problems or about the constructs of the programming language. In complex domains such as programming, tutors seem to engage in extensive reasoning about how to tutor students who make serious bugs. As an example of the kinds of issues a tutor reasons about when tutoring students in complex domains, consider the ostensibly simple problem of &en to deliver tutorial interventions. Two opposite strategies been proposed: A significant portion of tutorial interactions revolve around the bugs a student makes. When a tutor performs an intervention to help a student fix a programming bug, the problem of deciding which intervention to perform requires extensive reasoning. In this paper, we identify five tutorial considerations tutors appear to use when they reason about how to construct tutorial interventions for students’ bugs. Using data collected from human tutors working in the domain of introductory computer programming, we identify the knowledge tutors use when they reason about the five considerations and show that tutors are cottsistent in the ways that they use the kinds of knowledge to remon about students’ bugs. In this paper we illustrate our findings of tutorial consistency by showing that tutors are consistent in how they reason about bug criticality and bug categories. We suggest some implications of these empirical findings for the construction of intelligent tutoring systems.