Developing Shared Home Behavior Datasets to Advance HCI and Ubiquitous Computing Research



Researchers in fields such as HCI, product design, artificial intelligence, and telemedicine are increasingly interested in developing technologies for the home. Studying behavior in the home for purposes of design and engineering, however, can be logistically challenging and costly. Therefore, research in HCI-related fields may be accelerated by the creation of comprehensive, publicly accessible shared datasets on home behavior. These datasets may consist of recordings from sensors placed in homes, transcripts from interviews, notes from observations, or data collected using mobile devices or other means.

Shared behavioral datasets that include both quantitative sensor data and qualitative ethnographic data could be valuable to researchers from a variety of disciplines. For example, researchers interested in artificial intelligence may use portions of the dataset to test activity classification algorithms, while an interface designer might wish to review the various contexts in which a particular device is used. Meanwhile, a telemedicine researcher might work with the dataset to determine if a sensor placed on a specific object will provide enough information to monitor specific aspects of an individual’s health.

Ideally, researchers in HCI and ubiquitous computing would have access to datasets that could be used for exploratory algorithm and idea development, much as exists in other fields such as speech recognition, object recognition, and discourse understanding. These researchers would be able to conceive of and test ideas without expending resources to build and install sensor systems outside the laboratory. For some purposes, the datasets might need to include not only sensor and interview data, but also audio-visual recordings of the environment and annotations about ongoing activities.

Creating such datasets and the tools needed to efficiently share, browse, and collaboratively annotate them requires a major commitment of time and resources. This workshop is intended to identify priorities for efficiently developing such datasets and the tools to support them.

We seek applicants from a wide array of disciplines, including but not limited to: interface design, sensors, artificial intelligence, ubiquitous computing, ethnography, telemedicine, and behavioral psychology. Participants will be selected who can discuss their experiences of either creating, sharing, or using datasets on home behavior. We are also interested in hearing from people who may feel that shared, heavily used datasets are problematic for research (e.g., leading researchers to tune technology to the biases in the datasets).


The MIT organizers of this workshop are funded by National Science Foundation grant #0708375 (CRI: CRD Development of Longitudinal Home Activity Datasets as a Shared Resource).