Predictive Prototyping for Real-World Applications A model-based evaluation approach based on the ACT-R cognitive architecture

In this paper we explore the fusion of a sophisticated user interface prototyping tool with a cognitive modeling approach based on the ACT-R cognitive architecture. The resulting tool allows the creation of complex interactive prototypes for which quantitative performance predictions can be derived by running cognitive models. Such models are automatically generated by monitoring a designer's interaction while completing a task scenario with a prototype. Using the example of a real-world user interface for a software-controlled riveting machine, we demonstrate the scope of the approach and compare model predictions to data gathered in two empirical studies. The encouraging goodness-of-fit of the predictions strongly suggests that the use of such simulated interactions with prototypes provides a promising and cost-effective approach to predict learning trajectories and performance data for routine tasks. We argue that predictive prototyping fosters the acceleration of iteration cycles without resulting in additional expenditures for UX designers.