Cross-Device Consistency in Automatically Generated User Interfaces

The growing importance of ubiquitous computing has motivated an outburst of research on automatic generation of user interfaces for different devices (e.g., [6] or our own Supple [4]). In some cases, care is taken to ensure that similar functionality is rendered similarly across different applications on the same device [5]. However, we also need to ensure that after using an application on one device (say, a PDA) and having learned that user interface, the user will not have to expend much effort having to learn a brand-new user interface for the same application when moving to a new platform (e.g., a touch panel). We have began to extend our Supple system in a way that allows it to produce interfaces that make a trade off between optimality given a new platform and similarity to the previously rendered user interfaces for the same application. In particular:

• we show how to incorporate an interface dissimilarity metric into a UI generation process resulting in new interfaces resembling ones previously used by the user;
• we propose a list of most salient widget features that can be used to asses similarity of interfaces rendered on radically different platforms;
• and we outline the most promising approaches for automatically learning parameters of a UI dissimilarity function from user feedback.