Automated Partial Credit for STEM classes

Abstract

This paper introduces a system that allows the award of partial credit for exams and tests consisting of multiple choice and true/false problems. Traditional multiple-choice exams have a number of issues. For one, they result in overall lower grades than exams using the same problems in an open-ended format. It is well known, that students see multiple-choice and true/false problems as not completely fair because they do not take into account partially correct answers. Non-major courses such that Math and Statistics for Engineering and Computer Science majors create additional challenges because they require simultaneous content knowledge in more than one subject. A reliable automated method to assign partial credit for multiple-choice and true/false problems can greatly improve both quality and perception of fairness of the assessment. There are two major approaches to address the issue with multiple-choice problems, see e.g. [1], [2]. In this work we suggest a combined and generalized approach for assigning partial credit for non-open-ended problems by using Concept Inventory among other testing approaches. One of the main benefits of our suggested approach is content independence and applicability to both major and non-major STEM subjects.