Michael Madaio presentation

"Human-Centered Approaches to Fairness in AI"

As AI systems become increasingly ubiquitous in consumer technology and in high-stakes contexts, such as education, healthcare, and more, there is a growing awareness that AI systems may disproportionately harm marginalized groups. To address this, researchers have developed principles, toolkits, and metrics to support fairness in AI, yet these approaches are often divorced from the situated contexts in which AI systems are developed and used. In this talk, I will discuss two research projects with AI industry practitioners to co-design and use resources to support fairness in AI design. I will share insights into how these resources—an AI fairness checklist and a process for disaggregating AI model evaluations to uncover performance disparities—both shape and are shaped by the organizational contexts in which AI systems are produced. Finally, I will conclude with future directions for this research, including work in progress to understand how impacted communities and other stakeholders may meaningfully participate in the work of AI fairness.

Bio: Michael Madaio is a postdoctoral researcher at Microsoft Research working with the FATE research group (Fairness, Accountability, Transparency, and Ethics in AI). He works at the intersection of human-computer interaction and algorithmic fairness, focusing on enabling more fair and responsible AI through research with AI practitioners and people impacted by AI systems. Madaio received his PhD in Human-Computer Interaction from Carnegie Mellon University.