Telling people AI will destroy jobs undermines democratic trust and dampens political engagement: two preregistered, nationally representative experiments (replicated US/UK) and a Europe-wide survey show labor‑replacing frames erode support for democratic institutions and reduce willingness to participate in AI policymaking.

Main Finding

Across large-scale survey data and two preregistered experiments, people tend to perceive AI as more labor-replacing than labor-creating, and those perceptions (independent of measured technology-, political-, and sociodemographic covariates) are linked to lower satisfaction with democracy and reduced willingness to politically engage with AI. Experimental manipulation of framing (labor-replacing vs. labor-creating) causally reduced democratic trust and willingness to engage with AI policy.

Key Points

• Public perception: In a Europe-wide survey, the dominant public view frames AI as replacing jobs rather than creating them.
• Observational association: Perceiving AI as labor-replacing (vs. labor-creating) is associated with lower satisfaction with democracy and lower political engagement around technology, controlling for technology-related, political, and sociodemographic factors.
• Causal evidence: Two preregistered, nationally representative experiments (UK, N = 1,202; US replication, N = 1,200) show that exposure to a labor-replacing framing (vs. a labor-creating framing) produces greater erosion of trust in democratic institutions and reduces willingness to politically engage with future AI developments.
• Robustness/credibility: The experiments were preregistered and replicated across two countries, supporting external validity of the causal effect of framing.
• Perception versus reality: The political effects arise from perceptions and framing about AI’s labor-market impact, regardless of whether those perceptions match actual employment outcomes.

Data & Methods

• Cross-sectional survey: Large-scale dataset, N = 37,079 respondents across 38 European countries. Key variables included respondents’ belief about whether AI will on net replace or create jobs, measures of satisfaction with democracy, and self-reported political engagement with technology.
• Statistical controls: Analyses adjusted for technology-related characteristics, political variables, and sociodemographic covariates to isolate the association between perceptions and democratic attitudes/behavior.
• Experimental design: Two preregistered, nationally representative online experiments:
  • UK study: N = 1,202
  • US replication: N = 1,200 Participants were randomly assigned to read a vignette/frame emphasizing either AI as labor-replacing or AI as labor-creating. Primary outcomes were trust/satisfaction with democracy and willingness to engage politically around AI (e.g., contact officials, participate in debates or consultations).
• Outcomes: Both observational associations and experimental contrasts found statistically meaningful declines in democratic trust and political engagement under labor-replacing frames. (Exact effect sizes are reported in the original paper.)

Implications for AI Economics

• Political economy of automation: Public beliefs about AI’s labor-market effects can shape democratic legitimacy and political mobilization independently of actual labor-market outcomes. Economists modeling automation impacts should incorporate public perception as a feedback mechanism influencing policy adoption, regulation, and the pace of technological diffusion.
• Policy acceptance and implementation risk: If AI is predominantly framed (or perceived) as job-destroying, policymakers may face reduced public trust and lower civic participation in shaping AI policy, potentially increasing opposition to beneficial AI deployments or complicating governance.
• Communication and framing matter: How economists, firms, and policymakers communicate about AI’s labor impacts can materially affect public attitudes and engagement. Transparent, evidence-based messaging and attention to distributional consequences may mitigate negative political reactions.
• Mitigation strategies: Investment in visible policies that address perceived labor risks—retraining, income supports, job-creation incentives, sectoral transition plans—may preserve democratic legitimacy and encourage constructive public engagement with AI policy.
• Research priorities: Future work should link perceptions to actual labor-market outcomes over time, explore heterogeneity (by occupation, education, political ideology), quantify effect sizes on policy preferences and voting behavior, and test communication interventions that reduce misleading perceptions while promoting informed democratic participation.