Response Time to Detect Careless Responding and Its Relationship with and Prediction of Emotional Distress

Abstract

People experiencing emotional distress struggle with cognitive and motivational decline, which has been correlated with patterns of careless responding. Although several methods have been used to detect careless responses in emotionally distressed respondents, the response time has not been widely explored. The current study conducted secondary data analyses on a sample (N = 37,819) who completed the Depression Anxiety Stress Scale (DASS-42) in an online survey between 2017 and 2019. First, a response-time-based approach––a normative threshold method––was used to identify careless responding and examine its association with emotional distress using the DASS-42. Second, four machine learning models––decision tree (DT), random forest (RF), support vector machine (SVM), and naive Bayes (NB)––were trained on DASS-42 item responses and response times to predict emotional distress severity level. A significant correlation was found between the number of careless responses and subscale scores of anxiety and stress. In addition, Mann-Whitney U tests showed statistically significant differences between careless and careful responders in depression, anxiety, and stress. Regarding the machine learning models, SVM was found to be the best predictive model for classifying distressed people with an accuracy, sensitivity, and specificity exceeding 90%. Our results suggest that, in addition to survey responses, response time can identify careless responders and predict distressed responders.