Research - 04.11.2025 - 10:00
Past research suggests that automation (like robots replacing factory workers) often results in people feeling that technology could put their jobs at risk. In response, they start to demand more government help, such as better unemployment benefits, job guarantee or re-training programs. This is called the risk-insurance model of social policy preferences and according to this logic, people with low risk of losing their jobs should not support policies that benefit others because of the expense.
However, researcher Matthias Haslberger (University of St.Gallen) and his collaborators Jane Gingrich (University of Oxford) and Jasmine Bhatia (Birkbeck, University of London) wondered whether the recent rise of AI might change this pattern. The researchers noted that there are similarities and differences between AI and previous technological advancements that could have threatened jobs. AI has the potential to replace workers, but compared to industrial robots, AI affects different groups of workers and has arrived much more suddenly.
To gain insight into how people react when they are actually exposed to generative AI, the researchers conducted an online experiment with 1,041 working-age adults in the UK.
Participants were randomly divided into two groups. One group was given access to ChatGPT to help complete three realistic text-based work tasks, such as editing or evaluating conflicting arguments. The other group did the same tasks without AI. Afterwards, everyone answered questions about how risky they thought AI was for their job, their opinions on AI and what kinds of social policies they supported.
Surprisingly, exposure to AI did not make people more afraid of losing their jobs. The group that had access to ChatGPT didn’t feel more at risk of being replaced by technology or by someone better at using it. In fact, they felt more positive about AI in general—they saw it as helpful for themselves and society, and they were less likely to say the government should restrict its use.
But here’s the interesting twist: even though participants didn’t feel personally threatened, they were more likely to support government job guarantees and programs to help workers adapt or retrain. In other words, people didn’t see AI as a danger to themselves, but they still thought others might be harmed and wanted society to step in.
The researchers call this a “sociotropic” response—meaning that people care about how technology affects others and society as a whole, not just their own situation. Overall, people were cautiously optimistic, recognizing both the positive and the disruptive potential of generative AI.
This finding challenges the traditional risk-insurance model, which says that people demand government support when they personally feel threatened. Instead, at this early stage of AI adoption, people seem to support policies that protect others even if they themselves expect to benefit from AI – possibly because they recognise the disruptive potential of the technology for society.
The authors also note that these results reflect the stage of AI at the time. The study was conducted in mid-2023, when tools like ChatGPT were still new and not yet deeply integrated into workplaces. As AI gets more powerful and companies use it more widely, perceived personal job risk may rise, and the winners and losers will become clearer. In follow-up work, the researchers plan to study whether the “risk-insurance” logic might return or if people remain concerned about the effect of AI on others. “Self-interest is a powerful motivation,” explains Matthias Haslberger, “but our study shows that people also consider other factors. We need to better understand under which circumstances this is the case.”
The study Rage against the machine? Generative AI exposure, subjective risk, and policy preferences can be found here.
Matthias Haslberger is a postdoctoral researcher at the School of Economics and Political Science at the University of St.Gallen. His research focuses on the effects of artificial intelligence and technological change on labour markets and education systems.
Image: Adobe Stock / marumon
