Show Me How You Click, and I’ll Tell You What You Can: Predicting User Competence and Performance by Mouse Interaction Parameters

Abstract

Automatically detecting and adapting to user competence is a promising approach for advancing human-technology interaction. With the present work, we demonstrate that perceived user competence and performance can be predicted by easily ascertainable low-level mouse interaction parameters with considerable amounts of explained variance. N = 71 users with varying competence interacted with a statistical software while mouse interaction parameters were recorded. Results showed that perceived task competence could best be predicted by clicks per second, maximum mouse velocity, and average duration of pauses > 150 ms (R² = .39). Perceived system competence could best be predicted by clicks per second, maximum mouse acceleration, and average number of pauses > 150 ms (R² = .28). Performance could best be predicted by clicks per second, maximum mouse velocity, and average number of pauses > 150 ms (R² = .50). Results imply that assessing low-level mouse interaction parameters could be a feasible approach for automatic detection of user competence and performance.