Panel 1 Papers:
“Vetting the AI that Vets Us: Ethical Trade-offs in Public and Private Sector AI Adoption Decisions”
Daniel S. Schiff, Purdue University
Abstract: Artificial intelligence (AI) systems are increasingly used across the public and private sectors. Yet, little is known about which considerations shape decision-makers’ preferences around AI system adoption, use, and governance, nor whether public and private sector actors approach associated trade-offs similarly or differently. An especially critical area pertinent to a variety of public and private sector domains is human resource (HR) management. Drawing on a conjoint experiment and survey data from 1,459 HR professionals, 348 in government and 1,111 in the private sector, we examine how HR managers evaluate the relative importance of seven factors when making AI tool adoption decisions: AI system transparency, bias mitigation efforts, consent for impacted individuals, platform integration ease, data use and privacy considerations, human oversight, and cost. We find that decision-makers in both sectors take efficiency and ethical considerations seriously. However, public sector managers place greater weight on ethical issues like transparency, bias, human oversight, and privacy compared to their private sector counterparts. We also provide evidence on the state of AI adoption across various HR use cases, finding that HR managers view AI tools as relatively overused in certain high-risk settings but underutilized for more human-centered applications. These results suggest that relying on industry self-regulation alone may be insufficient to protect key public values.
“Explaining Women’s Skepticism toward Artificial Intelligence: The Role of Risk Orientation and Risk Exposure”
Peter John Loewen, Cornell University
Abstract: Rapid advances in artificial intelligence (AI) present substantial economic and social opportunities but also significant risks for different groups in society. Against this backdrop, this paper examines the gender gap in attitudes toward AI adoption, with a focus on how gender differences in risk orientation and risk perceptions drive skepticism toward AI’s economic benefits. Using original survey data from approximately 3,000 respondents across Canada and the United States, we find that women consistently perceive AI to be riskier than men. We identify two key drivers behind this gender gap: women’s higher general risk aversion and their greater exposure to AI-related risks. To establish a causal relationship between risk and AI attitudes, we further show experimentally that as the perceived benefits of AI become more uncertain, women’s support for companies adopting AI falls more sharply than men’s. Finally, structural topic modeling of open-ended responses confirms that women express greater uncertainty about AI’s benefits and more frequently anticipate little to no benefits. Given AI’s potential to exacerbate existing gender inequalities, our study highlights the critical importance of incorporating women’s perspectives into AI policy-making. Policies that do not address gender-specific risks may not only reinforce existing inequalities in employment and income but could also generate political backlash against AI adoption, reshaping political cleavages along gender lines.
“Robots Replacing Trade Unions: Novel Data and Evidence from Western Europe”
Paolo Agnolin, Princeton University
Abstract: Labor unions play a crucial role in liberal democracies by influencing labor market and political dynamics, organizing workers’ demands and linking them to parties. However, their importance has progressively diminished in the last decades. We suggest that technological change—and industrial robotization in particular—has contributed to weakening the role of unions. We produce novel granular data on union density at the sub-national and industry level for 15 countries of western Europe over 2002-2018. Employing these data, we estimate the impact of industrial robot adoption on unionization rates. We find that regions more exposed to automation experience a decrease in union density. The decline in unionization occurs via a compositional effect, i.e., a reallocation of employment away from traditionally unionized industries towards less unionized ones. On the other hand, there is no clear evidence of a systematic reduction in union density within industries more exposed to automation.
Panel 2 Papers:
“The Ubiquitous Word: How Information Revolutions Reshape Political Institutions”
Eduardo Albrecht, Columbia University
Abstract: This presentation examines humanity’s evolving relationship to knowledge through three transformative eras: the written word, the printed word, and today’s “ubiquitous word” of digital and AI-generated information. Drawing from arguments in my book Political Automation (Oxford UP, 2025) I describe how the age of information overabundance may strengthen rather than weaken democratic governance. We are experiencing a fundamental “gear shift moment” in not just how we relate to information, but more fundamentally in how we relate to knowledge about ourselves. This, in turn, changes the kind of political institutions we are capable of imagining—or willing to accept. Each information revolution has transformed political institutions in this way. The written word enabled massive religious bureaucracies; the printed word fractured them through reform and enlightenment. Today’s ubiquitous word represents a similar historic shift. As information becomes incredibly cheap to produce and disseminate throughout every level of society, the relation of individuals to knowledge about themselves also changes, challenging traditional asymmetries of power that have sustained political authority since the late-modern period. The presentation explores how this transition presents an opportunity for the evolution of democratic governance, and will examine how experimental institutional innovations around the globe may portend the emergence of a new form of direct and AI-augmented participation.
“Democracy in an AI-Augmented Bureaucracy”
Johannes Himmelreich, Syracuse University
Abstract: Democracies increasingly augment administrative decision-making with artificial intelligence (AI). This AI augmentation is a threat to democracy. Administrative decision-making supports democracy in various ways, but the mechanisms by which it does are insufficiently understood. Through case studies—from consumer protection, administrative statistics, and military contexts—I show how public servants contribute expertise, safeguard stability, resist political pressure, and hence serve the public interest. The case studies illustrate what I call “norms of responsible public service.” These norms not only explain how the AI augmentation of administrative decisions is a threat to democracy, these norms can also guide AI governance in evaluating and suggesting policies. Such policies may include stricter validation requirements for high-stakes applications, “job protections” for AI models to prevent swift overhauls by new administrations, and limits on automation in domains where human discretion remains essential.
“An Open Science System for Consistent and Persistent Measurement of Public Opinion”
Jason Jeffrey Jones, Stony Brook University
Abstract: Estimates of public opinion are most valuable when they are (1) timely, (2) consistently measured and (3) conform to open science expectations. Here, I present an automated system meeting all three requirements. Further, I present 18 months of results. In daily repeated surveys of a random sample of American adults, I measured attitudes toward Artificial Intelligence. Demographics, political party, generalized trust and risk willingness all predicted attitudes. Most prominently, however, attitudes changed over time - including an interaction between time and political party affiliation. This provides evidence that temporal validity is a challenge that deserves more consideration in the social sciences. The system I present here aims to meet that challenge.