CAEVA: cognitive architecture to evaluate visualization applications

We describe a cognitive architecture to evaluate visualization applications (CAEVA). This architecture requires gathering domain knowledge to adapt existing evaluation models to new domains and applications. This architecture has two main components: one for interoperability and one for cognitive models. The interoperability component uses client-server protocols. The cognitive component includes three subcomponents: knowledge about visualization, knowledge about the data domain, and knowledge about data analysis strategies. CAEVA was used to evaluate the visualizations generated to compare environmental products. The cognitive model is written in ACT-R and simulates how human beings use visualizations to perform data analysis tasks. The data produced by the simulation model is validated with empirical data obtained from human subjects performing the same task. The simulation model produces traces that allow researchers to analyze the strategies followed by the users in the task. The data produced by the model allowed us to study the learning process of visualization users.