Acceso usuarios
Noboru Matsuda, Evelyn Yarzebinski, Victoria Keiser, Rohan Raizada, William W. Cohen, Andreas J. Stylianides, Kenneth R. Koedinger
This article describes an advanced learning technology used to investigate hypotheses about learning by teaching. The proposed technology is an instance of a teachable agent, called SimStudent, that learns skills (e.g., for solving linear equations) from examples and from feedback on performance. SimStudent has been integrated into an online, gamelike environment in which students act as “tutors” and can interactively teach SimStudent by providing it with examples and feedback. We conducted 3 classroom “in vivo” studies to better understand how and when students learn (or fail to learn) by teaching. One of the strengths of interactive technologies is their ability to collect detailed process data on the nature and timing of student activities. The primary purpose of this article is to provide an in-depth analysis across 3 studies to understand the underlying cognitive and social factors that contribute to tutor learning by making connections between outcome and process data. The results show several key cognitive and social factors that are correlated with tutor learning. The accuracy of students’ responses (i.e., feedback and hints), the quality of students’ explanations during tutoring, and the appropriateness of tutoring strategy (i.e., problem selection) all positively affected SimStudent’s learning, which further positively affected students’ learning. The results suggest that implementing adaptive help for students on how to tutor and solve problems is a crucial component for successful learning by teaching.
