Responsible Governance of AI: Realities, Roadblocks, and Resolutions

I will begin by defining and advocating for the responsible governance of AI as crucial for maximizing the societal benefits of the technology while minimizing its harms. I will then discuss the key challenges faced by major AI actors and stakeholders in implementing responsible AI in practice, drawing from my work with practitioners and impacted communities. In particular, I will highlight the critical gap in the development of well-justified, valid, reliable, and feasible evaluation metrics, methods, and processes, identifying this as a significant barrier to operationalizing responsible AI, and describe how my work has aimed to address some of these issues using fairness measurement, AI red-teaming, and AI field-testing as case studies. Finally, I will outline an agenda for impact-driven research in this field, considering the broader socio-economic context and regulatory landscape.

Speaker: Hoda Heidari, Carnegie Mellon University