Direct manipulation involves a large number of interacting psychological mechanisms that make the performance of a given interface hard to predict on intuitive or informal grounds. This paper applies cognitive modeling to explain the subtle effects produced by using a keypad versus a touchscreen in a performance-critical laboratory task.

Dynamic behavior is enacted from static observations, but as experience is reconstructed with concepts, mutual exclusivity replaces correspondence between the factors of action and observation. This bistable reality generates incommensurable world views between groups, which constrain measurement and give rise to a tension that enables change but also results in misleading accounts, as if behavior were in response to virtual knowledge. In the form of independent scientific peer reviews at a Department of Energy site, tension was applied to improve practices of science in the field. Implications for a physics of interaction are addressed.

Increasing use of automation in computer systems, such as advanced cockpits, presents special challenges in the design of user interfaces. The challenge is particularly difficult when automation is intermittent because the interface must support smooth transitions from automated to manual mode. A theory of direct manipulation predicts that this interface style will smooth the transition. Interfaces were designed to test the prediction and to evaluate two aspects of direct manipulation, semantic distance and engagement. Empirical results supported the theoretical prediction and also showed that direct engagement can have some adverse effects on another concurrent manual task. Generalizations of our results to other complex systems are presented.