APIE Project: Personalized Appliance Control

Traditional remote controls typically allow users to activate functionality of a single device. Given that users activate a subset of functionality across devices to accomplish a particular task, it is attractive to consider a remote control directly supporting this behavior. We have conducted qualitative and quantitative research on two promising approaches to creating such a remote control: end-user programming and machine learning. In general, results show that each approach possesses advantages and disadvantages and that neither is optimal. Our next step is a build a mixed-initiative generator that draws on the strengths of both people and ML algorithms.

		An example of the current collection of devices and remotes that users must deal with.
		To determine how effectively users could create their own personalized remotes, we provided a variety of buttons and asked them to assemble the interfaces they wanted.
		Users create non-optimal remotes; they tend to create interfaces that center around devices rather than tasks, and they miss opportunities for macros (single buttons that issue multiple commands).
		Automatic clustering of commands into tasks, while facilitating the creation of interfaces that center around tasks rather than devices, requires a sufficient density of data before it yields high-quality results. This approach can therefore take too much time.

Publications

O. Omojokun, J. Pierce, C. Isbell, and P. Dewan. Comparing End-User and Intelligent Remote Control Interface Generation. 3rd Interactional Conference on Appliance Design, 2005. A version of this paper will appear in Personal and Ubiquitous Computing.

C. Isbell, O. Omojokun and J. Pierce. From Devices to Tasks: Automatic Task Prediction for Personalized Appliance Control. In Proceedings of 2AD, 2004. A version of this paper appeared in Personal and Ubiquitous Computing.