Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This work examines how to detect, document, and prevent plagiarism in exams. First, to identify and quantify plagiarism, we propose methods that compare similarities in multiple-choice answers between seat neighbors and nonneighbors. Second, we document cheating in undergraduate exams. Under baseline monitoring, at least 7.7 percent of the row-wise neighbor pairs plagiarized. Pairs composed of academically weaker students cheated more. Third, using a field experiment, we demonstrate that close monitoring eliminated cheating. By contrast, signing an honesty declaration doubled cheating relative to the control group. Complementary experiments suggest that the declaration backfired because it weakened the social norm of academic integrity.