An ACT-R Model of Elliptical Endpoint Error Distributions in a Mobile Touchscreen 2-D Fitts Law Task
Kristen Greene, Melissa A. Gallagher, Franklin Tamborello
Given the high propensity of users motoric errors with smaller touchscreen usa-buttons, knowing the endpoint distributions for finger- based pointing and tapping is especially important for higher-fidelity predictive modeling of tasks on such devices. One of the most studied models of aimed human motor movement in HCI is Fitts Law. While Fitts Law has a long and successful history of application in predicting mouse-pointing or stylus-tapping times, its traditionally high predictive ability declines when it is applied to finger-pointing tasks involving small touchscreen targets, especially when the finger (input device) is larger than the target itself (commonly known as the fat finger problem). There is still some uncertainty regarding the systematic prediction of endpoint distributions for two-dimensional finger-pointing tasks. Recent work (May, 2012) found endpoint error distributions larger for on- axis than off-axis movement in a mouse-pointing task, with the shape of the error distribution along the movement axis more ovoid than circular around the target center. Since the implementation of endpoint error in ACT-R did not previously distinguish between on-axis and off-axis error, Gallagher and Byrne (2013) implemented Mays 2012 work in an ACT-R model by modifying the method by which noise is added to the ending position of mouse movements. Here, we build upon such modifications by implementing them in an ACT-Touch model; ACT-Touch is an extension to the ACT-R cognitive modeling framework, useful for modeling and simulation of human interactions with mobile touchscreen devices.