Modeling Touch Input for Users with Motor Impairments: Empirical Insights into Training Size Requirements

Abstract: We present empirical insights into modeling touch input performance, in terms of time and offset to targets, using Gaussian models and a cross-validation framework based on prediction coverage and bias of z-scores with log-likelihood analysis. Our results, from data collected from seven participants with upper-body motor impairments, indicate an optimum window of 8 to 24 touch observations before models plateau in performance. We interpret our findings through the lens of ability-based design, and propose a four-step procedure—observe, model, revise, and share—for implementing touch targets adaptive to users’ motor abilities. The procedure is lightweight, requiring only basic numerical computations of touch time and offset measurements available on all platforms, making it readily deployable across a variety of touchscreen devices.

Conference: CHI EA ’26: Proceedings of the Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems, April 13 – 17, 2026

Publication: Association for Computing Machinery, New York, NY,  United States

Link: https://dl.acm.org/doi/proceedings/10.1145/3772363