An untapped area of productivity gains hinges on automatic detection of user cognitive characteristics. One such characteristic, spatial visualization ability, relates to users' computer performance. In this dissertation, we describe a novel, behavior-based, spatial visualization detection technique. The technique does not depend on sensors or knowledge of the environment and can be adopted on generic computers. In a Census Bureau location-based address verification task, detection rates exceeded 80% and approached 90%.

In the field of Human Computer Interaction and Psychology, it is accepted that spatial visualization (VZ) is one ability that can indicate individual’s performance on computer applications. Since users with different levels of VZ seem to prefer different types of user interfaces (UI), knowing a user’s level of VZ provides a great opportunity for application developers to design software with higher satisfaction and usability. In this paper, we proposed three models to predict a participant’s level of VZ based on the participant’s actions (taps) on the tablet screen while doing an address verification task in the neighbor- hood using the tablet. After applying the proposed prediction models with data of thirty participants, they yielded an optimal accuracy of 93.33%.