Cognitive Demand of Model Tracing Tutor Tasks: Conceptualizing and Predicting How Deeply Students Engage

Model tracing tutors represent a technology designed to mimic key elements of one-on-one human tutoring. We examine the situations in which such supportive computer technologies may devolve into mindless student work with little conceptual understanding or student development. To analyze the support of student intellectual work in the model tracing tutor case, we adapt a cognitive demand framework that has been previously applied with success to teacher-guided mathematics classrooms. This framework is then tested against think-aloud data from students using a model tracing tutor designed to teach proportional reasoning skills in the context of robotics movement planning problems. Individual tutor tasks are coded for designed level of cognitive demand and compared to students’ enacted level of cognitive demand. In general, designed levels predicted how students enacted the tasks. However, just as in classrooms, student enactment was often at lower levels of demand than designed. Several contextual design features were associated with this decline. Implications for intelligent tutoring system design and research are discussed.