Some participants of online surveys engage in extreme answering behavior while generating responses (i.e., they respond too fast or too slow) relative to population norms. Here, we demonstrate how participants’ navigation behaviors can be used to potentially identify such responses. We administered an online survey where students (who were earlier instructed to complete a task) report lenience scores towards non-appropriate behavior while completing the task. We draw on cognitive dissonance theory to posit that failure to follow instruction predicts lenience scores. We then created different datasets by excluding data from participants flagged by our metrics and generated predictive models. We found that model performance improves by removing data from flagged participants, indicating a reduction in noise from the dataset. Despite demonstrating the effectiveness of our approach, we encourage researchers to exercise caution and elaborate on the limitations of our approach and future avenues of research.
Kumar, Manasvi; Valacich, Joseph; Jenkins, Jeffrey; and Kim, David, "Too Fast? Too Slow? A Novel Approach for Identifying Extreme Response Behavior in Online Surveys" (2022). SIGHCI 2022 Proceedings. 4.