Optimal Weighting for Exam Composition
This addresses the challenge for instructors in creating more accurate exams, though it appears incremental as it builds on existing weighting methods with algorithmic optimization.
The authors tackled the problem of designing exams that accurately assess student abilities by developing a novel framework that uses machine learning algorithms to optimize exam question weights, using overall class grades as a proxy for true ability, and they achieved significant error reduction compared to standard weighting schemes.
A problem faced by many instructors is that of designing exams that accurately assess the abilities of the students. Typically these exams are prepared several days in advance, and generic question scores are used based on rough approximation of the question difficulty and length. For example, for a recent class taught by the author, there were 30 multiple choice questions worth 3 points, 15 true/false with explanation questions worth 4 points, and 5 analytical exercises worth 10 points. We describe a novel framework where algorithms from machine learning are used to modify the exam question weights in order to optimize the exam scores, using the overall class grade as a proxy for a student's true ability. We show that significant error reduction can be obtained by our approach over standard weighting schemes, and we make several new observations regarding the properties of the "good" and "bad" exam questions that can have impact on the design of improved future evaluation methods.