Mediation Analysis of Workload and Emotion on the Performance of Supervision Tasks

Understanding how users’ emotional responses and cognitive workload affect task performance is essential for designing effective systems. Rather than treating emotion and workload as responses, this study examines whether they can statistically mediate the relationship between task conditions and task success. The analysis used the open-source MOCAS dataset in which 21 participants monitored robot swarms across 9 conditions varying in robot speed and camera count. A multilevel, multivariate mediation framework was applied with three continuous mediators (arousal, valence, workload) and one continuous outcome (success rate). Task variables significantly increased workload and arousal while decreasing valence ( p  p  < .05), although only valence showed significant mediator-specific indirect effects. Task variables also produced significant negative direct effects on performance, especially at higher robot speeds. Overall, findings suggest that valence is a potential statistical mediator of task-performance than workload or arousal alone in human supervision tasks.