Intelligent Tutoring Systems show promise for improving the efficiency of teaching by tailoring instruction to each student's needs. This research proposes two innovations over current tutoring methods. The first, is interest based, i.e., instruction based on the basis of a model of the student's interest. The second, is the use of multiple intelligent agents to deliver individualized instruction and engage in collaborative problem solving with the student. The end use of the research will include commercial games that teach scientific and mathematical domains.
