In Progress

Working Papers

• Balancing Progress and Protection: Do Citizens Want Governments to Steer Technological Adoption? Conditionally accepted

  With Alexander Kuo.

  Explains when citizens support government intervention to steer, slow or regulate technological adoption.

  Abstract

  The rapid progression of the fourth industrial revolution has sparked debate about its economic impacts, and prompted calls for increased government "steering" (through regulation or taxation) of technology adoption. This study proposes that preferences for such policies are shaped by two main considerations: the need to protect certain workers from technological disruption increases support for government steering, but the association of technology with economic growth and consumer benefits reduces such support. To test theories of such preferences, we design novel survey questions about six steering policies, measure preferences in large, representative samples from five European countries, and experimentally vary the claims presented. We also measure many technological risks, including job substitution by artificial intelligence (AI). We find substantial support for government steering of technology adoption through regulation and taxation, but claims that such intervention harms consumers or growth substantively diminishes such support, while arguments about protection enhance support only modestly.

• What If You See It? Workers' Perceptions of and Reactions to Generative AI. Under review

  With Massimo Anelli, Italo Colantone and Piero Stanig.

  Shows how seeing generative AI perform one's own job tasks can polarize workers' policy and political reactions.

  Abstract

  How do workers perceive Generative AI? More as an opportunity or as a threat for their occupational prospects? And how do these perceptions influence their policy and political preferences? We conducted a survey experiment on around 6,000 workers in 98 occupations directly exposed to Generative AI, in Germany, Italy, and the US. Treated respondents are shown a video of ChatGPT performing the most frequent core task in their occupation. On average, they become more optimistic about the technology's impact. Yet, a substantial share of treated respondents become more threatened. This leads to polarization of policy preferences: threatened individuals increase their support for technology regulation and redistribution policies, while optimistic individuals move the other way. The threatened also express warmer attitudes toward "backlash" political parties.

• Building Global Support for AI Governance: Evidence from Six Countries. Under review

  With Alexander Kuo and Shir Raviv.

  Identifies which institutional designs make international AI governance more acceptable to publics across major economies.

  Abstract

  The rapid advancement of AI presents unprecedented challenges requiring international coordination, yet efforts to establish global governance frameworks remain fragmented. An important yet understudied factor shaping governments' willingness to negotiate, approve, and enforce international agreements is domestic support. We examine how key institutional features of AI governance shape public support for international arrangements using new data from a large-scale survey experiment conducted in the US, China, India, Germany, the UK, and Japan. Our results indicate that citizens worldwide are willing to accept—and even prefer—inclusive, enforceable, and neutrally led regulations, suggesting that an ambitious global governance framework of AI could receive broad public support. The findings highlight governance mechanisms that generate global support and areas of cross-national disagreement that may require tailored approaches.

• Defaulting to the State: Accountability Concerns, Instrument Uncertainty, and Demand for AI Regulation. Draft available

  With Alexander Kuo and Shir Raviv.

  Examines why public demand for strict AI regulation often reflects accountability concerns and uncertainty about policy tools.

  Abstract

  Recent surveys reveal strong public support for stringent regulation of AI, yet we still know little about what this demand means for citizens. We argue that mass support for strict AI regulation partly reflects a combination of concerns about who is accountable about AI-related risks and lack of knowledge about how oversight could be organized. Using novel data from six countries that together account for 75% of world investment in AI, we find broad support for government oversight across diverse AI applications, yet open-ended responses reveal that only a minority connect this support to specific policy instruments. Experimental evidence shows that learning about non-ban oversight mechanisms progressively softens demands, suggesting that what citizens demand is credible governance in a domain where the available instruments remain poorly understood. To conclude, we discuss the implications of our findings for the analysis of public opinion in new or technical policy domains.

Other Work in Progress

• Perceptions of AI as an Opportunity or a Threat: The Role of Resources and Context. Draft available

  With Alexander Kuo, Erick Padilla and Florencia Olivares.

  Studies which resources make workers more likely to see AI as an opportunity rather than a threat.

  Abstract

  Do workers view AI as a threat or an opportunity for their jobs, and do these views matter for policies regulating AI in the workplace? The literature has established that objective occupational exposure to new technologies correlates, albeit weakly, with subjective concerns. This paper argues that, at equal levels of exposure to AI, three types of resources shape whether individuals embrace new technologies with optimism or concern; these are cognitive, non-cognitive, and place-based resources. We test these claims using a new survey of 34,212 respondents across 15 European countries. We find that education and psychological confidence are strongly correlated with AI occupational "optimism," while the hypothesized regional-contextual effects are not. In addition, we show that perceptions of AI as an opportunity or a threat in turn shape support for government intervention on this topic. To conclude, we discuss the implications of our findings for policy in times of rapid labor market change.

• Digital Skills Training and Political Integration: A Field Experiment with Public Employment Services in Catalonia. In progress

  With Florencia Olivares. Pre-analysis plan available on request.

Recent Funded Projects