Until recently, automation was mostly a concern for workers in factories or routine clerical jobs. The assumption was that knowledge workers — lawyers, journalists, analysts, academics — were largely insulated from technological displacement. Generative AI has unsettled that assumption. It can now perform tasks across a wide range of knowledge-intensive work, and the anxiety about job replacement has spread accordingly.
In a new article published in Perspectives on Politics, I examine how workers across 27 countries are responding to this and what kinds of policies they want in response. The findings complicate some long-standing assumptions about welfare politics.
The limits of existing welfare paradigms
Modern welfare states in rich democracies have largely been organized around two logics. The first is social consumption: unemployment benefits and income support for people who lose jobs or face hardship. The second is social investment: education and retraining to help workers stay employable as economies change.
Both rest on assumptions that AI is beginning to strain. Social consumption has increasingly been tied to work and prior contributions, on the logic of encouraging labor market participation. That becomes awkward when entire occupations, not just individual workers, are transformed or displaced. Social investment, meanwhile, is built on the premise that new technology mostly raises the demand for higher skills, an assumption that is harder to sustain when AI is threatening tasks performed by highly educated workers.
What the data show
The article draws on the 2024 OECD ‘Risks that Matter’ survey. Around 35% of workers said it was likely or very likely that AI would take their job or job opportunities within the next five years. Notably, high-skilled workers are now as worried as low- and middle-skilled workers — an important shift from earlier waves of this survey.
The policy preferences are revealing: support for expanding unemployment benefits sits at around 50% and is no higher among workers who feel threatened by AI than among those who don’t. Retraining programs remain broadly popular, but support is actually lower among those who feel most directly exposed. Workers who fear AI replacement appear less convinced that the existing paradigms — compensation after the fact, or reskilling for new jobs — are adequate to what they face.
The appetite for something different
The clearest finding concerns two policies that sit outside the established welfare paradigms. Support for universal basic income rises among workers who feel threatened, reflecting a desire to reduce dependence on paid employment rather than simply restore it. But the strongest result is for robot taxes: levies designed not primarily to raise revenue, but to reduce firms‘ financial incentives to automate, thereby slowing the pace of technological transformation. Among workers who feel threatened by AI, support reaches around 60%, making it the most preferred response. This holds across countries, regardless of existing welfare generosity.
A different political logic
What these findings suggest is that workers who feel exposed to AI are not simply asking for more of what welfare states already provide. They are expressing something closer to a demand for earlier intervention: for shaping the terms of AI deployment before disruption occurs, rather than managing its consequences afterwards.
This represents a different political logic. It treats the pace and direction of AI adoption as an object of collective decision-making, not an inevitability to which social policy must simply adapt. The question it raises is not primarily how societies should support workers through technological change, but whether the transformation of work is itself something that can and should be governed democratically.
The article does not claim to resolve these questions. The data show associations, not causal effects, and public attitudes toward new technologies can shift. But the patterns are consistent enough to suggest that citizens are already connecting AI with broader questions about work, welfare, and the responsibilities of the state, and that the politics emerging around these questions may not fit neatly into existing frameworks.
Protecting workers after disruption may no longer be sufficient. The impact on work is itself a political question, and workers are starting to treat it as one.
Read the full article: https://doi.org/10.1017/S1537592726104538
