Statistical Modeling of Cognitive Fatigue Through Digital Writing Dynamics: An Exploratory Multivariate Regression Study

Digital fatigue in higher education is associated with sustained interaction with digital interfaces and may affect academic performance. This study evaluates the feasibility of a non-intrusive detection approach based on keystroke dynamics, avoiding cameras and physiological sensors. A dataset of N = 85 natural writing sessions was analyzed using temporal metrics (dwell time and flight time), typing speed, and backspace-based error rate. Descriptive results indicate strongly nonnormal and heavy-tailed temporal distributions, particularly for flight time. Correlational analyses reveal weak associations between perceived fatigue (treated as approximately continuous) and timing variables. A simple linear regression model identifies error rate as a statistically significant predictor, explaining 11.8% of variance (R 2 = 0.118) in fatigue scores. These findings support the feasibility of keyboard-based digital fatigue monitoring in naturalistic academic settings while highlighting substantial inter-individual variability [1, 8, 9].