Artificial Intelligence and Democratic Freedoms

Civic Technology Community Group,

Hello. I'm pleased to share with the group the following hyperlink to a recent symposium, Artificial Intelligence and Democratic Freedoms: https://knightcolumbia.org/events/artificial-intelligence-and-democratic-freedoms . At the webpage, videos are available with respect to each of the following panels and keynote addresses. Each panel also has papers' abstracts available.


Panel 1: Regulating AI in a Time of Democratic Upheaval

Liberal democracies had almost reached consensus on how to regulate narrow AI when ChatGPT came along and general purpose AI took center stage. How should regulatory proposals adapt to this new reality—and what prospects are there for effective regulation when the global geopolitical order is being reconfigured?

Panel 2: AI Agents’ Democratic and Economic Impacts

The most consequential societal impacts of large language models will come not through the content they generate but through the cognitive resources they provide to effective software agents. This panel considers how these agents in turn will reconfigure political and economic realities.

Keynote Address and Q&A

Ben Buchanan, Johns Hopkins University

Panel 3: Evaluation and Design of Safe AI

Advances in large language models were enabled by, and fostered, a boom in technical AI safety research. But narrowly technical approaches not only mischaracterize AI risks, they miss key opportunities for mitigating them. This panel fosters a more multidisciplinary, sociotechnical approach to technical AI safety.

Panel 4: Using AI to Enhance Democracy

Democratic freedoms are imperiled by affective polarization and radicalization. This panel explores how AI can, and cannot, be used to forge agreement or even consensus in formal and informal democratic deliberation.

Panel 5: Reframing AI in the Digital Public Sphere

Democracies depend on a vibrant public sphere. No social institution has been more shaped by narrow AI than the digital public sphere—by most accounts, to its detriment. This panel asks if more capable general purpose AI systems can do better.

Panel 6: AI, Freedom of Expression, and Democratic Backsliding

As large language models become more and more integrated into the information environment, alignment techniques designed to limit AI companies’ exposure to liability and public embarrassment are affecting the expressive interests of model users (including jailbreakers). At the same time, the community of research on AI risks is facing unprecedented political intervention into the content and methods of its research. All this is happening at a moment of rising authoritarianism and democratic backsliding. This panel asks how to square the values of safe AI with core principles of freedom of expression—preserving expressive interests against both private and state control.

Keynote Address and Q&A

Erica Chenoweth, Harvard University

Panel 7: Transitions to Transformative AI

Even if AI capabilities stalled out at their present level, the societal ramifications would be massive. But if progress continues at roughly the same pace, we may be dealing with even more profoundly transformative AI systems within the decade. This panel seeks to get on the front foot—to characterize what ‘transformative AI’ would mean, how it would differ from what has been realized so far, and how to develop societal resilience to the potential for further radical social change.



Best regards,
Adam Sobieski