The first two cognitive tutors we developed in the mid 1980s were a Lisp Programming Tutor and a Geometry Proof Tutor. Each of these tutors is a problem-solving environment analogous to the Algebra I Cognitive Tutor displayed in Fig. 9.1. The Lisp Tutor, which is closely integrated with a Lisp programming text (Anderson, Corbett, & Reiser, 1987), has been in continuous use in a self-paced programming course at Carnegie Mellon University since 1984. Students read through the text and complete corresponding programming tasks with the help of the Lisp Tutor. The Lisp Tutor has proven to be an extremely productive research environment (Anderson, Conrad, & Corbett, 1989; Corbett
& Anderson, 1995; Corbett & Trask, 2000) as well as a highly efficient learning environment. Students working with the Lisp Tutor completed programming problems in as little as one-third the time required by comparable students working in a conventional programming environment, as displayed in Fig. 9.3, while scoring 25% higher on subsequent tests (Corbett & Anderson, 1991).
The Geometry Proof Tutor (GPT), which was developed to support students in completing Euclidean proofs, was piloted in a Pittsburgh high school from 1985 to 1987.
This project served as an early prototype for our current high school math tutor project.
Students used the tutor in their regularly scheduled geometry classes and with the geometry teacher in the room. GPT also proved to be a highly effective learning environment. Students in the GPT classes scored a letter grade higher on a subsequent paper-and-pencil geometry proof test than comparable students in control classes, who spent the same amount of time in conventional classroom problem-solving activities, paper-and-pencil seatwork, and boardwork.
During this pilot, Janet Schofield, a social psychologist at the University of Pittsburgh, completed an important observational study of GPT’s impact on the classroom as part of a larger study of the impact of computer technology in a Pittsburgh high school (Schofield, 1995). She noted that the cognitive tutor transformed the classroom in two general ways, as summarized in Fig. 9.4.
Schofield observed that cognitive tutors transformed the teacher-student relationship.
Teachers spent more time interacting with students who needed the most help, in contrast with whole-class instruction in which teachers tend to interact with the more successful students. The teachers spent the most time in extended interactions with individual students in the act of problem-solving and learning by doing. Schofield also documented that students found cognitive tutors motivating. She noted increased time spent on tasks, greater involvement, and increased effort. She noted that students in the comparison classes spent as much as 15 minutes chatting about nonacademic topics. In contrast, during the tutor sessions, it was common for students to begin working before the starting bell and to continue working after the closing bell. In one anecdote, she noted that a fist- fight almost broke out between a student who arrived early to work on geometry proofs and another student who was working late on the same workstation.
Technology Enhanced Learning 205
FIG. 9.3. Average Lisp programming problem completion times across five lessons for students using the Lisp Tutor and students working in a conventional programming environment.
Both the Lisp Tutor and Geometry Proof Tutor represent hothouse successes. The self- paced course that employs the Lisp Tutor (and subsequent Prolog and Pascal tutors) has been taught exclusively by its developers. When GPT was piloted a member of the university research team was always present in the classroom. To take the next step, these two projects were followed by the ANGLE Geometry Tutor Project, which marked a transitional phase in moving from the research lab to the classroom (Koedinger &
Anderson, 1993). Like GPT, ANGLE is a problem-solving environment in which students construct graphical representations of Euclidean proofs, as shown in Fig. 9.5.
206 Cognitive Tutors: From the Research Classroom to All Classrooms
FIG. 9.4. Impact of cognitive tutor technology on the classroom (Schofield 1995).
In each problem-solving step, the students select premises, post a conclusion, and identify the theorem justifying the conclusion. As with the earlier tutors, ANGLE proved to be a highly effective learning environment, but even more importantly the ANGLE project provided two lessons that helped guide the subsequent Algebra I project. For the first time, we encountered a curriculum compatibility problem. By the time ANGLE was piloted, the city high schools had adopted a new geometry text that de-emphasized proofs (in response to the 1989 National Council of Teachers of Mathematics curriculum standards), and it was difficult to integrate the tutor into the course. Second, the ANGLE evaluation study revealed a significant teacher interaction effect. The project teacher who had helped develop the tutor and was intimately familiar with the integration strategy achieved greater learning effects than other teachers who were much less familiar with the integration strategy. This result underscores the critical role of addressing the context surrounding educational technology use. Without careful attention to curriculum integration and sufficient training for teachers, an otherwise good solution may not reach its potential.
Technology Enhanced Learning 207
FIG. 9.5. The ANGLE Geometry Tutor. Reprinted from A.Corbett, K. Koedinger, and J.Anderson, Intelligent Tutoring Systems, copyright © 1997, p. 862., with permission from Elsevier Science.