Effective Feedback Content for Tutoring Complex Skills

Feedback during learning is critical for evaluating new skills. Computerbased tutoring systems have the potential to detect errors and to guide students by providing informative feedback, but few studies have evaluated the real impact of different types of feedback. This article presents results of such a study using the Geometry Tutor for building geometry proofs. It was found that feedback about the goal structure of geometry problems led to better performance than feedback about the reasons for error or than simply being told that an error had occurred. This goal feedback allows students to correct the incorrect action more often than other types of feedback. Also, the goal feedback group continued to deal advantageously with problems when the feedback was subsequently removed. A simulation model, based on Anderson's (1983) ACT* theory and an analogical learning system, presents a preliminary model of the effects of these different feedback types. The model indicates that the advantage of goal-directed feedback is a reflection of its immediate application to the problem, whereas feedback about reasons for the error does not provide any direction to the correct error. According to the model, the feedback allows the student to construct a correct representation of the goal tree involved in various types of proofs more readily than feedback that is not immediately relevant to the current problem. Author's present address: Jean McKendree, Microelectronics and Computer Technology Corporation Human Interface Lab, 3500 West Balcones Center Drive, Austin, TX 78759.