A Testbed for Teaching Problem Solving Skills in an Interactive Learning Environment

An interactive learning environment was developed with the goal of empirically testing the effectiveness of various teaching strategies in improving problem solving performance. The domain chosen was transfusion medicine since it involves solving complex, multiple solution problems which are typically found to be difficult (Elstein, Shulman, and Sprafka, 1978) and because normal performance of this task calls for marking data sheets with intermediate conclusions, thereby improving the chances of the computer correctly inferring the student's reasoning. The testbed, called TMT (for Transfusion Medicine Tutor), monitors for errors, builds a model of what a student knows and can select teaching strategies based on human tutoring models that were developed from earlier studies. The testbed will be used to collect data of a student's performance in conditions where the degree of teaching and type of feedback are manipulated. A number of broadly applicable issues can be explored in this framework such as the difference between expert and student problem solving strategies, the effectiveness of different teaching strategies, and the importance of modeling student knowledge and providing visual feedback when developing an interactive learning environment. Preliminary results of our experiments, a demonstration of the testbed, and a discussion of how it was implemented will be presented in the demonstration session.