RANDOMIZED EXAMS: LEVERAGING UNCERTAINTY TO IMPROVE STUDENT LEARNING

Abstract

This thesis explores the impact of randomization of exam timing on student effort in a dynamic principal-agent model. The model com pares between deterministic exams, which have predetermined dates, and randomized exams, where students are uncertain about when they will be tested. Based on standard assumptions regarding the shape of students’ cost functions, the analysis finds that randomized tests result in significantly more effort than deterministic tests. This out come arises from the need for the student to maintain some prepared ness under uncertainty. Additionally, the study outlines optimal grad ing functions, demonstrating that a simple pass/fail exam successfully leverages the benefits of random timing, while a two-threshold method can optimize performance by encouraging effort more flexibly through out the periods. The findings presented contribute to the existing lit erature on monitoring, contract design, and information asymmetry by demonstrating that uncertainty over timing - when combined with well-designed grading standards - can be used as a means to improve agent performance in an educational environment.