Cheating Detection in Tests: A Systematic Review

Primary author: Tarid Wongvorachan
Faculty aponsor: Chad Gotch

Primary college/unit: College of Education
Campus: Pullman

Abstract:

High-stakes testing is significant in education across the world. Unfortunately, widespread cheating undermines the interpretations and uses of test results. When test-takers cheat, other test-takers, testing programs, and test users can suffer substantial negative consequences. Educational measurement professionals have developed numerous cheating detection methods to counter these potential consequences, but to-date no research has undertaken a comprehensive inventory of the field. The purpose of this systematic review is to document current trends and identify needs for further research, in order to improve the security of high-stakes testing programs. From an exhaustive library database search, I selected 62 primary studies for the in-depth review. Each study was reviewed for both general (e.g., authorship) and specific level characteristics (e.g., application of empirical data and type-I error). This review produced 27 variables that were synthesized to portray characteristics of the field as a whole.

Examination of these variables showed that cheating detection methods are in an emerging stage. Future research needs to expand beyond the current focus on western countries, employ more real test data (vs. simulated), and use more varied data sets. The field also needs a single standard to assess proposed detection methods. Further, machine learning could be a viable addition to the predominant statistical approaches observed in the literature. Expansion of the research base in these directions could help strengthen the security of high-stake testing in education, and ultimately support valid interpretations and uses of test scores.