As AI changes how people access, do their jobs, and interact with government systems, what will be the political consequences? And if AI is to be used in public decision-making and policy implementation, how can it be governed in a way that preserves public trust and democratic accountability?
These are the questions that have driven postdoctoral researcher and political scientist Shir Raviv during her two years at the Data Science Institute.
“Ensuring that the design and use of these tools reflect citizens’ values and the public interest is crucial for maintaining legitimacy and democratic accountability,” Raviv says. “A key part of my research agenda is to uncover what these values are and how they can be effectively incorporated into AI governance frameworks.”
At Columbia, Raviv has worked under the mentorship of Jeffrey Fagan, Isidor and Seville Sulbacher Professor of Law and Professor of Epidemiology, Tamar Mitts, Assistant Professor of International and Public Affairs, and Jason Healey, Adjunct Professor on International Affairs and Senior Research Scholar at the School of International and Public Affairs.
She designs field experiments in which people interact with AI-driven systems in real-world work settings, as well as large-scale surveys that track how public attitudes shift based on experience and information exposure.
“I use experimental methods to understand people’s attitudes towards high stakes algorithmic decision-making,” Raviv says. “Under what kinds of conditions are these decisions acceptable? How might their views change if they are exposed to additional information or have personal experience interacting with the technology?”
One strand of her research explores how online workers respond when supervised by algorithmic systems. Another uses national surveys to identify what makes algorithmic decisions seem fair, legitimate, or accurate. She has also examined the influence of media coverage of AI on public opinion in the U.S., and conducted a multinational study examining attitudes towards A.I. governance in the US, China, India, Germany, the UK, and Japan.
She found consistent patterns. Across domains, Americans strongly oppose reliance on AI in decision-making about punishment, as well as when algorithms are used to make inferences about individuals rather than groups. Her field studies with online workers revealed that while first hand experience with AI can influence workplace behavior, it has limited impact on broader attitudes toward AI policy.
In contrast, information from trusted, nonpartisan sources about AI’s ethical, economic, or political implications can significantly shift public views, suggesting that cross-cutting coalitions around public-interest regulation could be effective in shaping future regulation. Similarly, her multinational study found broad support for inclusive, enforceable, and neutrally administered AI regulation, a hopeful sign for a future global governance framework.
Raviv is now developing this research into a book, Governing (by) Algorithms: The Mass Politics of AI in Policy Implementation, that examines the key elements that shape public attitudes towards algorithmic decision-making systems in public policy. Her work highlights the need to understand and incorporate societal attitudes into the governance of AI systems – especially in high-stakes areas like criminal justice, policing, welfare, and education.
“Many initiatives to govern AI overlook the views of citizens–those who live with the consequences of algorithmic decisions but have no option to opt out,” she says. “This gap can, and already has, triggered public backlash. Even well-designed systems could fail politically if they don’t reflect public values.”
Raviv, who will join the faculty of Tel Aviv University in the fall, reflects on how interdisciplinary work has been critical to her research, especially in a field where ambitious ideas for responsible AI fall short in practice.
“There’s often a gap between the aspirational goals of researchers and designers of responsible AI principles and the technical challenges of achieving them in practice,” Raviv noted. “The interdisciplinary environment at DSI allowed me to refine my research to be more responsive to these challenges. It provided me with both the practical and applied perspectives that are essential to my work.”