Understanding expert-novice differences in geometry problem-solving tasks: a sensor-based approach

Understanding learner differences with sensors is increasingly important for effective learner modeling. Learner models based on a student's problem-solving actions and the automated interpretation of those actions have successfully advanced computer tutoring services. However, such transaction level actions provide insufficient detail about higher-rate cognitive variations, which may hold key information about individual differences in cognition and learning, and about factors that differentiate attention-switching strategies and instructional effects between individuals. To fill this gap, we have conducted a user study to investigate causal relationships between learners' expertise levels and patterns of interaction and attention during learning tasks by using an eye tracker and physiological sensors. In this paper, we validate our experimental test-bed built for inferring learners' cognitive processing states and diagnosing learning phases with sensors, and present initial results about expert-novice differences revealed in transaction level samples and sensor data streams.