Power, Equity, and Agency in Technology Design

At the time of our study, image search results for occupations underrepresented women and exaggerated gender stereotypes; the first woman in Google's results for "CEO" was CEO Barbie. Working with Matthew Kay and Cynthia Matuszek, we documented this pattern across occupations and showed that skewed results can shift people's beliefs about who works in a field. The paper became an early empirical foundation for research on algorithmic bias, is widely cited in HCI and machine learning fairness research, and drew coverage from The New York Times, The Washington Post, The Atlantic, and the BBC.

This work connects to my earlier research on diversity in news aggregation, where we designed algorithms and interfaces that expose people to a wider range of political viewpoints, and studied how people respond to that exposure.

Autoplay, infinite scroll, and frictionless defaults, sometimes called dark patterns, maximize engagement at the expense of people's ability to make intentional choices about their time and attention. Kai Lukoff led our study of how YouTube's design shapes users' sense of agency, and we found that common platform features undermined the agency users said they wanted. We then built and evaluated SwitchTube, a proof-of-concept system with "adaptable commitment interfaces" that let users set their own constraints in advance.

Companies market smart home cameras for security and peace of mind, but people use them in more varied ways. Neilly Tan led our study of everyday camera use, in which we found that the same devices people used to watch pets and check on packages also let them observe neighbors, often without the neighbors' knowledge or consent. In a companion speculative design study, we explored how people imagine and negotiate ethical boundaries for cameras in the home. Together, these studies show how products designed around individual use can produce collective harms, and they raise questions about what designers owe the people their products capture.

Wearable trackers have spread quickly in college athletics, where coaches, athletic departments, and universities, not athletes, usually decide to deploy them. Sam Kolovson led our study of this practice, in which we found that data collected from athletes' bodies primarily served institutional interests and that athletes had little control over how it was used or who saw it. In a follow-up speculative design study, athletes, coaches, and administrators described divergent visions for the future of sports tracking that reflect these power asymmetries. This work informs questions about consent and data governance when tracking is a condition of participation, and it connects to my research on personal informatics.

Health data that supports self-management can also be used against the people who collect it. After the 2022 Dobbs decision overturned Roe v. Wade, Georgia Kenderova led our study of how people navigated the new risks of period tracking and other health technologies. Most participants did not substantially change their tracking practices despite privacy concerns. They weighed abstract legal threats against concrete benefits and switching costs, and some saw tracking as both a risk and a protection. We argue that designing for sensitive health data requires attending to the legal and political contexts in which privacy features operate, not only to the features themselves.

Questions of power and agency extend to the workers who build technology. After the 2022–23 tech layoffs, Sam So led our longitudinal study with 29 laid-off tech workers. Participants described alienation and unfulfillment, along with a conflicted attachment to tech work that we characterize, drawing on Lauren Berlant, as “cruel optimism”: they remained oriented toward an industry that no longer offered what drew them to it, while struggling to imagine or reach alternatives. The study, covered by GeekWire and KUOW, connects individual workers’ experiences to broader questions about labor, resistance, and how tech work is imagined and valued.

HCI researchers increasingly work with people who face structural marginalization, and this work raises ethical questions without clean answers. Calvin Liang led our study of how 24 HCI researchers navigate this work, many of whom experience marginalization themselves. We identified four tensions: exploitation (researchers gain far more from studies than participants do, and marginalized researchers can be tokenized by colleagues), membership (shared identity with participants helps but guarantees nothing, and should not decide who may do the research), disclosure (requests for positionality statements can pressure marginalized researchers to out themselves), and allyship (much of what is called allyship is performative). We argue that researchers should embrace these tensions rather than expect to resolve them, and we propose an allyship-oriented approach to research with marginalized people, along with reflective questions for authors and reviewers weighing disclosure. Researchers across many populations and design contexts have since taken up the framework.