In contrast to the vast amount of modeling work focused on desktop user interfaces, recent work has increasingly focused on “off-the-desktop” interfaces, one prime example being in-vehicle interfaces used while driving. This paper highlights four recent approaches to predicting driver distraction from in-vehicle interfaces as secondary tasks: hand-crafted modeling with the full-fledged ACT-R architecture, handcrafted modeling with the much less complex ACT-Simple framework, modeling-by-demonstration using the new CogTool, and simplified modeling-by-demonstration using the integrated Distract-R system. While all four use an integrated-model approach and a rigorous driver model, each approach illustrates different advantages and disadvantages of simplifying cognitive modeling for purposes of rapid prototyping and evaluation of in-vehicle interfaces. (Author/publisher)
Abstract