My research program is focused on understanding and finding solutions to problems that arise in sociotechnical systems as they become an increasingly invisible and indispensable part of everyday life. A sociotechnical system involves people, technology, and information; these parts all interact and influence each other, and without all three parts the system would not function as it should. These systems have great potential to help people and improve their lives; however, they also have the potential for harm.
In particular, I focus on sociotechnical systems that are “black boxes” from the perspective of people using the system—the inputs and outputs can be observed, but the inner workings can’t be and are therefore hard for people to reason about. A social dilemma arises when machine learning algorithms make new inferences from datasets containing information collected about large numbers of people over time. The inferences, or “derived data”, can be used to provide beneficial functionality for people using these systems. But, the same data might also be used to infer sensitive personal information that people might want to keep to themselves.
Right now, the systems people use give them few opportunities to learn about, let alone manage, the types of inferences and predictions that are being made using data they and others have provided. This means that people can’t even begin to reason about consequences to themselves and others, because they don’t have any way of knowing what inferences are possible. In my work, I’m trying to find ways to help people regain some agency with respect to the platforms they use, by creating opportunities that do not exacerbate information overload for people to be exposed to data and potential inferences.
My research is currently funded by NSF SaTC award CNS-1524296, and by an endowment to MSU from AT&T. Some keywords to describe my research interests are: algorithmic curation, personalization, automation, digital privacy, derived data, ubiquitous computing, sociotechnical systems, user-contributed content, and human computer interaction.