A Cognitive Model for Automatic Student Assessment: Classification of Errors in Engineering Dynamics

The present paper focuses on the errors that students made on a first year engineering dynamics final exam. An error classification scheme, based on Action Theory, is used to classify errors as either mistakes or slips in logic. Simple rules are then developed for a computer to be capable of categorizing errors based on a priori indicators such as: the student's experience level and the topic's priority on the courses concept inventory as well as a posteriori indicators such as the frequency of the error within the exam. The assessment algorithm is then trained and optimized. The resulting algorithm is tested by comparing the results with a second, independent, expert. Results of the study are then discussed.