In this dissertation, I show that many appliance usability problems can be addressed by mov-ing the user interface from the appliance to a handheld device that the user is already carrying, such as a Personal Digital Assistant (PDA) or mobile phone. This approach, called the Personal Universal Controller (PUC), takes advantage of the increasing pervasiveness of wireless communication technologies that will allow handheld devices to communicate di-rectly with appliances. Automatic generation of the appliance user interface allows the PUC to create interfaces that are customized to the platform of the controller device, the user's previous experience, and all the appliances that are present in the user's current environment.

This dissertation makes several contributions to the state of the art in automatic interface generation:

• Automatic generation that makes use of dependency information to determine a better structure for the generated user interfaces
• The general Smart Templates technique for incorporating domain-specific design con-ventions into an appliance specification and automatically rendering the conventions appropriately on different platforms and in different interface modalities
• Algorithms that apply knowledge of similarity between specifications and interfaces the user has previously seen to generate new interfaces that are personally consistent
• Algorithms that use a model of the content flow between appliances to generate task-based interfaces that combine functionality from multiple appliances

An evaluation of the PUC system compared the automatically generated interfaces for two all-in-one printers with the manufacturer's interfaces for the same two appliances and found that users of the automatically generated interfaces were twice as fast and four times as suc-cessful for both common and complex tasks. The evaluation also shows that the PUC's consistency features allow users to be twice as fast when using a new appliance that is similar to an appliance they have previously encountered. This evaluation is the first known user study of automatically generated interfaces compared to human designs, and it shows that automatic generation of user interfaces for end users is now viable for interactive systems.

358 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu