Analyzing Stress Responses Related to Usability of User Interfaces

This paper presents a pilot study conducted to explore the potential relationship between estimated stress levels based on physiological signals and the usability of an interactive system. To this aim, we performed a set of usability tests with 30 participants, in which the physiological signals were recorded using the Empatica E4 wristband. The collected data were analyzed using the Stress.io app, which used a regression model trained on the AffectiveROAD dataset to predict the stress level from the wristband’s physiological data. As a pilot study, we performed a usability test on a main common task of an interactive system: the registration to a website. Users, divided into two groups, tested two different versions of the interface: the first group used a "bad-designed" one that violated most of the Nielsen heuristics while the second group tested a "well-designed" version in which all the usability issues were solved. During the performance of the task, each participant wore the E4 wristband and its sensors data were collected for further analysis. Following the tests, participants were requested to complete the System Usability Scale (SUS) questionnaire to evaluate the interface’s usability. Additionally, participants’ emotional reactions were captured using the Self-Assessment Manikin (SAM), which assessed valence, arousal, and dominance dimensions. The collected sensor data were automatically analyzed using the developed stress prediction model. The analysis of results yielded interesting findings. The standard deviation of estimated stress values exhibited an inverse correlation with the SUS score, indicating that greater variability in stress levels corresponded to lower perceived usability. Moreover, positive correlations were observed between the valence and dominance of reported emotional reactions and the SUS score, while an inverse correlation was found with the standard deviation of stress levels. Conversely, arousal reported in the SAM displayed inverse correlations with both the SUS score and stress level standard deviation. These outcomes suggest that participants reporting negative emotional reactions and higher arousal tended to perceive lower usability and experience heightened variability in stress levels.