ABSTRACT
In the late 1970s several researchers had the insight that artificial intelligence might be productively grafted into computer-based instructional environments. Reports of these now-classic projects in the book Intelligent Tutoring Systems (Sleeman & Brown, 1982) inspired our research team to begin developing a type of intelligent learning environment called “Cognitive Tutors™.” Cognitive tutors are rich problem-solving environments constructed around a cognitive model of the learner. They are designed to facilitate learning by doing, to make thinking visible, and to support complex problem analysis, solution and communication. Our primary motivation 15 years ago was to develop authentic learning environments in which to test the evolving ACT-R cognitive theory (Anderson & Lebiere, 1998), but it quickly became clear that the cognitive tutors are very effective learning environments (Anderson, Corbett, Koedinger, & Pelletier, 1995). Students working with our early programming and geometry proof tutors outperformed comparable students in conventional learning environments by one standard deviation. This is about half the effect that human tutors can achieve (Bloom, 1984), but two or three times larger than the average effect of conventional computer-assisted instruction (e.g., Cohen, Kulik & Kulik, 1982; Kulik, 1994; Niemjec & Walberg, 1987).
