Representing Student Performance with Partial Credit

The educational data mining community has not been paying much attention to how much assistance a student needs. Feng and Heffernen[1] showed that we can predict student performance better by accounting for amount of assistance, but didn't reduce it to a value that could be shared with students. In this paper we want to see if we can better model student performance by replacing traditional binary measures (correct/ incorrect) with continuous partial credit, which is assigned based on the details in the student responses. The data we use is from ASSISTment, a web-based math tutoring system for 7th-12th grade students. The system helps the student learn the required knowledge by breaking the problem into sub-questions called ‘scaffolding’ or by giving the student hints on how to solve the question.