Events

November 12, 2026

AI Auditing Past, Present, and Futures

1:30pm CDT,  Center for Human-Computer Interaction + Design (Frances Searle Building 1-122)

We stand at a pivotal moment in the history of AI auditing: audits are about to become routine and standardized, with implications for the world, and simultaneously, a little further on the horizon they’re also about to fundamentally change shape, shifting from point-in-time evaluations to ongoing, longitudinal monitoring.  

 AI auditing is a method developed over roughly the last decade, adapting the audit study (a social science technique used since the mid-20th century to detect discrimination in housing and hiring) to black-box algorithmic systems as the latter became used to make similar decisions. It involves systematically probing a system by submitting controlled inputs and then measuring whether outputs differ across groups or conditions, to draw inferences about how the system works and its impacts on users and society. Auditing has proved effective in surfacing discrimination and other consequential harms in areas like ad delivery and facial recognition systems, prompting companies to change their products, fueling public debate, and informing legal action. In the last few years, auditing has also been taken up and codified in laws and regulations around the U.S. As a result, auditing is moving from an academic and journalistic practice into a compliance function, which raises new questions about who conducts audits, against what standards, and to what end. Simultaneously, cutting-edge research in AI auditing is developing the method in a new direction, towards live, ongoing monitoring.  

Danaé Metaxa is the Raj and Neera Singh Term Assistant Professor of Computer and Information Science  at the University of Pennsylvania, with a secondary appointment in the Annenberg School for Communication. Dr. Metaxa's research focuses on the responsible deployment of AI in contexts including employment, politics, and advertising. This work includes auditing and evaluating algorithmic systems for bias, fairness, and accountability, alongside human-centered interventions for identifying and proposing algorithmic improvements. Dr. Metaxa holds a PhD in Computer Science from Stanford University, and has had research published and awarded in top-tier publication venues including ACM CHI and CSCW, and most recently is an author of MIT Press’ Essential Knowledge Series book, Auditing AI. Dr. Metaxa has also worked in an advisory role with legal, industry, and policy groups including the NYC Commission on Human Rights, the American Civil Liberties Union (ACLU), and PolicyLink, and serves as UPenn’s delegate to the U.S. National Institute of Standards and Technology’s AI Safety Institute Consortium. 

May 14, 2026

AI-Resilient Interfaces for Sensemaking and Beyond

1:00pm CDT,  Center for Human-Computer Interaction + Design (Frances Searle Building 1-122)

AI is powerful, but it can be wrong, contextually inappropriate, or subjectively bad—and suppress the user’s critical engagement and learning at the same time. My group is leading a charge to define and refine our understanding of AI-resilient interfaces that help people use AI while (a) maintaining their ability to notice AI choices, e.g., by spurring curiosity and cognitive engagement through surfacing variation in the data and/or the AI’s outputs and (b) giving them the context necessary to evaluate whether those choices are wrong, or at least not right for them. This is critical during context- and preference-dominated open-ended tasks, like ideating, searching, sensemaking, and reading or writing text and code at scale. AI-resilient interfaces improve AI safety, usability, and utility by working with, not against, human perception, attention, and cognition. To achieve this, we are deriving design implications from cognitive science and empirical studies of novel system prototypes, even and especially when these implications fly in the face of common usability guidelines.

Elena L. Glassman is an Assistant Professor of Computer Science at the Harvard John A. Paulson School of Engineering & Applied Sciences, specializing in human-computer interaction. Prior to that, she was a postdoctoral scholar at UC Berkeley, and obtained a BS, MEng, and PhD in Electrical Engineering and Computer Science from MIT. She has been named a Stanley A. Marks & William H. Marks Professor at the Radcliffe Institute for Advanced Study and a National Academy of Sciences Kavli Fellow. Her work has been funded by the NSF, private industry, the Berkeley Institute for Data Science, and the Sloan Research Fellowship. This work has received Best Paper and Honorable Mention awards at top-tier human-computer interaction research venues.

April 30, 2026

(Re)Working AI: Labor, Design, and the Governance of Workplace Technologies

1:00 pm CDT,  Center for Human-Computer Interaction + Design (Frances Searle Building 1-122)

Artificial intelligence is frequently introduced into workplaces with the promise that it will make work easier, safer, or more efficient. In practice, many automated systems only function because workers intervene when those systems fall short. This talk examines how organizations deploy workplace technologies in ways that depend on frontline intervention while restricting workers’ influence over how those systems are designed, evaluated, and governed. Drawing on ethnographic and design research in waste management, hospitality, and public transportation, the talk shows how AI systems impose standardized logics on labor that is inherently contingent, embodied, and relational. When those logics break down—or conflict with the demands of safety, service, or bodily limits—workers absorb the consequences: adapting robotic systems, navigating algorithmic schedules, and exercising discretion under tightly constrained operating conditions. This work is essential to keeping organizations functioning, yet it is rarely recognized as expertise or incorporated into decisions about procurement, deployment, or oversight. The talk then traces what changes when worker expertise is taken seriously as a basis for governance. Across cases, I examine labor-aligned design work, collective bargaining, and policy interventions that shift decision-making power—ranging from co-design efforts that surface worker knowledge, to procurement constraints, oversight requirements, and regulatory obligations that make that knowledge institutionally binding. A consistent pattern emerges: worker participation becomes consequential only when it alters who has the authority to set limits, define acceptable risk, and halt or redirect deployment. Taken together, these cases show that AI governance depends on whether working people have meaningful authority in the creation, implementation, and governance of the technologies that organize their labor.

Sarah Fox is an Assistant Professor at Carnegie Mellon University in the Human-Computer Interaction Institute, where she directs the Tech Solidarity Lab. Her research examines how AI, automation, and algorithmic management systems are designed, deployed, and governed in everyday work settings, and how these processes redistribute power between institutions, technologies, and workers. Drawing on ethnographic fieldwork and design research, she foregrounds workers’ situated expertise as a critical site of governance, tracing how everyday practices of adaptation, contestation, and repair expose the limits of technological systems and inform mechanisms of accountability and enforcement.

March 19, 2026

Redefining the Image Captioning Challenge: From Truthful to Useful

1:00pm CDT,  Center for Human-Computer Interaction + Design (Frances Searle Building 1-122)

Visual media increasingly dominates modern (online) communication, but this heavily visual landscape creates substantial barriers for those who cannot access it. To facilitate non-visual accessibility, we need to translate visual information into linguistic descriptions. What appears straightforward on the surface is in fact a fundamental communicative challenge: when moving between modalities, what exact information should be selected and conveyed? While modern vision–language models achieve ostensibly “superhuman” benchmark results on image-to-text tasks, they frequently underperform in real communicative settings. I argue that this gap reflects a deeper problem of information selection: effective image descriptions require not just stating what is literally true about an image, but deciding what should be said given a specific communicative goal and audience. Drawing on cognitive experiments with sighted participants and people who are blind or have low vision, I show that communicative goals strongly shape which visual information people choose to express, and that current AI tasks and evaluation methods fail to capture this nuance. I conclude by discussing recent work in which we start to quantify informativity in image descriptions, and how information selection varies across cultures, highlighting a broader research agenda for understanding and modeling cross-modal communication.

Elisa Kreiss’ work combines tools from natural language processing, psycholinguistics, and human-computer interaction to advance our understanding of how communicative context shapes language use. This research has direct applications to image accessibility – the challenge of (automatically) generating image descriptions for blind and low vision users.

November 6, 2025

Digital Culture Shock: Who Creates Technology and Why This Matters

1:30pm CDT,  Center for Human-Computer Interaction + Design (Frances Searle Building 1-122)

Robots that encroach on your personal space, baffling emojis, a chatbot that gives you an answer that seems terribly rude—does any of this sound familiar? If so, you may know what it feels like to experience a clash of cultures in technology. In this talk, I will present key insights from my book Digital Culture Shock: Who Creates Technology and Why This Matters (Princeton, 2025), showing how culture—shared values, norms, and behaviors—influences both the design of technology and its use. Drawing on a set of common assumptions that do not translate across cultures, the talk will outline what is at stake and how we can resist generalizing our own cultural peccadillos in technology design.

Katharina Reinecke is a Professor and Associate Director for Research and Communication in the Paul G. Allen School of Computer Science & Engineering at the University of Washington, Seattle. She has been speaking and writing about intelligent user interfaces that seamlessly adapt to people’s cultural and demographic backgrounds for nearly two decades and has pioneered ways for reliable, large-scale data collection from participants around the world with her virtual lab, the LabintheWild. Katharina received a PhD in Computer Science from the University of Zurich and was a postdoctoral fellow at Harvard University. Prior to coming to the University of Washington, she was an Assistant Professor in the School of Information at the University of Michigan.