Students in science courses often receive far too little guidance to allow reliable learning of needed reasoning skills. As a result, many students fail to acquire flexibly usable scientific knowledge (even with modern well-designed instruction and materials). Furthermore, students often lack some basic cognitive abilities needed for complex intellectual tasks (e.g., making well-considered decisions or adequately assessing their own performance). The goals of this project are to improve science instruction by the following approach: 1. Use cognitive analyses to specify the procedural knowledge needed for effectively applying scientific concepts or principles and for solving problems. 2. Help students learn such procedural knowledge, teach them explicitly the basic cognitive functions of deciding on an action, implementing it, and assessing the result. 3. Do this by adapting a "reciprocal-teaching strategy" (very effective in teaching reading) where a student and tutor trade roles, alternately (a) deciding and assessing or (b) implementing. 4. Develop novel computer programs ("PALs", i.e., Personal Assistants for Learning) that can play the role of the tutor--so that students and computers coach each other. Such PALs, which involve careful instructional design but no artificial intelligence, can provide every student with good individual guidance and feedback. (Preliminary experiments suggest that such PALs might achieve instructional effectiveness approaching that of expert human tutors.) This work is expected to lead to the following results: 1. Production of PALs of demonstrated instructional efficacy and utility in basic physics. 2. Effective methods for developing PALs, in physics as well as in other domains. 3. Increased understanding of relevant cognitive and human-computer interaction processes.
