Simulating Human Cursor Trajectories for Path-Sensitive GUI Evaluation

Abstract

User simulation models support predictive evaluation of interface designs by generating synthetic interaction behavior. While existing models can accurately simulate unconstrained pointing and clicking, many common graphical user interface interactions impose geometric constraints on cursor motion, where the path taken directly affects interaction outcomes. We present a parametrizable generative user simulation model that generates realistic cursor trajectories for such tasks by formulating constrained movement as a receding-horizon optimization problem using Model Predictive Contouring Control. The model balances speed, smoothness, and continuous path and boundary compliance. Evaluation against human data in both abstract tunnel steering tasks and realistic interface scenarios, including cascading menus and lasso selection, shows that the simulated trajectories closely match human behavior and support scalable, trajectory-level analysis of path-sensitive interface designs.