Providing students with online warning messages about potential cheating during an assessment reduces a student's cheating behavior for the remainder of the assessment.
This study included 1,077 undergraduate students at a Chinese university who were participating in a high-stakes exam. Cheating was measured by two metrics: 1) students exiting their full screen exam mode, and 2) through trained coders identifying “cheating” behaviors via webcam. Students were given either a control message or a warning message after completing one-third of the exam. Notably, before receiving the message the control and the warning groups did not vary significantly in cheating behaviors. However, after receiving the warning message, the warning group displayed significantly fewer cheating behaviors, as compared with the control group. The cheaters' test scores compared to noncheaters had the largest variations, such that cheaters who received the control message had the highest scores (possibly inflated due to cheating) and cheaters who received the warning message had the lowest test scores. These findings suggest that providing students with warnings during the assessment is effective in reducing subsequent cheating behavior. Other research, in a similar context, noted that warning students before an assessment had no effect on cheating behavior. Taken together, warning students during compared to before an assessment appears to be more effective at reducing cheating.
The effectiveness of warning messages that are delivered at the beginning of a test may wear off over time and students may become accustomed to these messages if delivered consistently.
Students who were identified as cheaters, based on video monitoring and jump-out responses, displayed significantly less cheating behavior after receiving a warning middle warning message, as compared to students who were not identified as cheaters.
Students who are engaged in cheating behavior will likely stop after receiving a middle warning message, while students who are not engaged in cheating behavior will not be influenced to a statistically significant degree.