Predictive Cognitive Modelling of Applications

This paper argues that important usability aspects of mobile applications can be automatically evaluated using computational cognitive models based on the cognitive architecture ACT-R. A tool incorporating cognitive models for specific tasks, users, applications and usability aspects is proposed. Explanations provided by the tool for usability flaws are based on simulations of cognitive mechanisms. A use-case of the tool is introduced, which is based on an ACT-R model that simulates how users search and select a specific target in a hierarchical android application and predicts efficiency and learnability for average users. The model has been empirically validated in four studies with two different applications. To fully automate the usability evaluation of the use-case, two basic requirements need to be fulfilled. First, the application and the cognitive model have to be connected. A tool called ACT-Droid acts as an interface between the Android application and the cognitive model. Second, the models knowledge of the world, which is application specific, has to be provided automatically by using an automated user interface analysation approach. Therefore, the open-source tool AppCrawler was extended to allow the extraction of the required information.