Alix Dunn from Ada Lovelace Institute talks about FAccT conference and papers presented. Topics include AI model marketplaces moderation, algorithmic bias in policing, AI auditing laws, responsibility in AI supply chains, disgorgement of data for models, and surveillance concerns for AI technology.
FAccT conference focuses on AI ethics, with research on algorithmic biases, responsible AI practices, and system abandonment.
Moderating AI model marketplaces is crucial to prevent misuse and ensure legal compliance.
Diverse stakeholder perceptions on bias in predictive policing tools challenge reformist and non-reformist approaches.
Integration of academic research with policymaking, engaging diverse communities, and addressing military AI ethics are essential.
Deep dives
Exploring the Fairness, Accountability, and Transparency Conference
A conference named FACT focused on AI and algorithmic technologies and their impact on fairness and accountability was discussed. The conference gathered social scientists, data scientists, engineers, and policy experts to explore the coexistence of technology and society, substantial research, community engagement, and interdisciplinary discussions.
Moderating AI Model Marketplaces
A paper by Michael Veal and Robert Gorawa studied companies like Hugging Face and GitHub that host AI models. They addressed the challenges of moderating open-source models on these platforms due to potential misuse and legal concerns, emphasizing the need to evaluate and moderate model marketplaces to prevent unlawful uses or harm.
Algorithmic Bias in Predictive Policing
The paper by Marta Ziosa and Dasha Perus examined algorithmic bias in predictive policing tools, revealing how bias perceptions varied among different stakeholders like police officers and the community. The research highlighted complex views on bias and showcased reformist and non-reformist reformers' perspectives on policing approaches.
Challenges in Implementing Responsible AI Practices
A research by Michael Modello at Google and David Witter from Cornell delved into the challenges of operationalizing responsible AI practices within companies. They highlighted the need for clear guidance, organizational change, and structural support to ensure a uniform understanding and implementation of ethical AI practices.
Addressing Technology Abandonment
Research on abandoned algorithmic systems analyzed cases of system abandonment due to ineffectiveness in areas like care provision algorithms. The study identified phases of abandonment like discovery, diagnosis, dissemination, dialogue, decision, and death, reflecting the decision-making and impact of discontinuing AI systems.
Policy Implications and Future Directions
The podcast conversation explored various policy considerations, community engagement, research translation into actionable policies, and the importance of bridging academia, activism, and policy to drive meaningful societal changes. It highlighted the need for creating a research agenda that informs policy decisions and addresses pressing socio-technical challenges.
FTC's Regulatory Approach and AI Ethics
The insights shared by Stephanie Nguyen, CTO at the FTC, shed light on the FTC's consumer protection regulation strategies, including remedies like deletion of data and disgorgement for AI systems created using ill-gotten information. The discussion emphasized regulatory powers, governance structures, and the responsible use of AI technologies in alignment with ethical principles.
Closing Remarks and Reflections on AI Ethics Discourse
The final segment discussed the integration of academic research with policymaking agendas, the cultivation of a more engaging and inclusive academic community, and the need for intellectual abstraction to build bridges between different stakeholders in the AI ethics space. It also highlighted the urgency of addressing military applications of AI and navigating the complex ethics and ideologies surrounding AI technologies.
In part 1 of our FAccT conference deep dive, Alix Dunn sits down with co-host Andrew Strait from the Ada Lovelace Institute to talk about the history of FAccT and some of the papers being presented at this year’s event.
The Fairness, Accountability and Transparency Conference, or FAccT is an interdisciplinary conference dedicated to bringing together a diverse community of scholars and exploring how socio-technical systems could be built in a way that is compatible with a fair society. The seventh annual FAccT conference was held in Rio de Janeiro, Brazil, from Monday, June 3rd through Thursday, June 6th 2024 with over five hundred people in attendance.
Lara Groves (Ada Lovelace Institute, UK), Jacob Metcalf (Data & Society Research Institute, USA), Alayna Kennedy (Independent researcher, USA), Briana Vecchione (Data & Society Research Institute, USA) and Andrew Strait (Ada Lovelace Institute, UK)- Auditing Work: Exploring the New York City algorithmic bias audit regime
Nari Johnson (Carnegie Mellon University, USA), Sanika Moharana (Carnegie Mellon University, USA), Christina Harrington (Carnegie Mellon University, USA), Nazanin Andalibi (University of Michigan, USA), Hoda Heidari (Carnegie Mellon University, USA) and Motahhare Eslami (Carnegie Mellon University, USA)- The Fall of an Algorithm: Characterizing the Dynamics Toward Abandonment
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
