| Touchscreens have seen widespread adoption in the last decade due to the rise of smartphones, tablets, and touchscreen laptops. New interaction techniques have emerged that match the capabilities of touch interfaces, with advantages for usability, but one problem has worsened: the probability that an error will occur in the basic task of choosing interface elements has significantly increased over point-and-click interactions. The limits of the human motor system have been documented extensively, and fingers are not precise input devices. Our approach is to develop cognitive models of users' actions in this context. Computational modeling is a staple in cognitive science, an interdisciplinary field that integrates theories from psychology, physiology, and computation, in order to explore human cognition.;Researchers in human-computer interaction increasingly see the value of being able to predict user performance on a given task, by use of a model, before conducting user studies.Models derived from observed user behavior can also shed light on the cognitive strategies and microstrategies that may generalize to other domains, a topic of interest among cognitive scientists.;This dissertation makes three contributions. First, we present an extension to Cogulator, an open source cognitive modeling program, that allows for probabilistic branching and looping behavior in GOMS models. Second, we used that extension to identify and model the low level cognitive strategies that users apply in recovering from touch-screen targetselection errors. Cognitive modelers interested in touchscreen interactions can use this information to better simulate real human performance. Lastly, we developed interaction guidelines for visual feedback that can assist users in recovering more quickly. |
