ARCHIVES

Cognitive workload monitoring is essential for improving human–computer interaction, productivity, and user well-being in modern digital environments. Traditional workload assessment techniques often rely on intrusive physiological sensors or self-report measures, which can disrupt natural user behaviour and limit real-time applicability. This study presents the design and implementation of a real-time, non-intrusive machine learning system for cognitive workload detection using mouse and keyboard dynamics. The proposed system continuously captures low-level interaction features such as keystroke timing, mouse movement patterns, click frequency, and cursor trajectories during normal computer usage. These features are processed and fed into supervised machine learning models to classify users' cognitive workload levels in real time. Experimental evaluations demonstrate that interaction-based behavioural features can effectively reflect variations in cognitive demand, achieving promising classification accuracy without requiring additional hardware or explicit user input. The system offers a scalable, cost-effective, and privacy-conscious solution for continuous cognitive workload assessment, with potential applications in adaptive user interfaces, workplace productivity monitoring, e-learning platforms, and human-centred computing systems.