Online surveys are used for collecting self-report data. Despite their prevalent use, data quality problems persist due to various response biases. Here, we demonstrate how participant answering behaviors can be used to identify biased responses. We administered an online survey where participants reported their personality dimensions of neuroticism and extraversion—two personality dimensions that have been previously shown to be correlated with a propensity to deceive—and were later presented with a scenario to exhibit deceptive behavior. We then generated models to predict deception using the neuroticism and extraversion constructs. Using respondents’ fine-grained mouse movement data when answering these questions, we generated time, behavior, and navigation-based metrics to identify biased participants. By removing these outliers, model performance improved by 93% for neuroticism and 10% for extraversion. This approach aids in gaining a clearer understanding of how some types of response biases influence model performance