Detecting emotional stress during typing task with time pressure

Automated stress detection is important in developing adaptive e-learning systems. Empirical evidence suggests that mouse dynamics and keyboard dynamics analyses can be both effective in user behaviour modelling as well as emotion detection compared to physiological measures and facial expression recognitions, and yet they are far less expensive and considered non-intrusive. This paper investigates how mouse dynamics and keyboard dynamics can be affected by emotional stress, particularly stress induced by time pressure, text length and language familiarity. Our research findings show that longer text and unfamiliar language raise users' stress perceptions. Demanding job such as long typing task could result in anomalous behaviours once the users have lost motivation. Language familiarity mainly affects keyboard behaviour but text length change mouse behaviour. This shows that there is a good potential of developing an adaptive e-learning system by detecting learners' emotional stress from keyboard and mouse input.