Survey puts public estimate of human extinction risk near 5%, with modest urgency

Researchers led by Matthew Coleman and Lucius Caviola published a survey of lay beliefs about human extinction, drawing on five studies and 2,147 participants in the United States and China. The paper, titled “Lay beliefs about the badness, likelihood, and importance of human extinction,” addressed extinction risk broadly rather than AI-caused extinction specifically, though it sits inside a debate — over how much weight the public gives extinction scenarios that AI safety advocates raise — that runs directly through AI policy discussion.

The headline finding was a gap between perceived probability and perceived priority. US respondents gave a median estimate of around 5% for the likelihood of human extinction this century, but said the probability would need to reach roughly 30% before they would call preventing it society’s single highest priority. Respondents did view extinction prevention as one among several issues deserving significantly more resources, but not the dominant one, and the researchers found these priority judgements were largely resistant to change from reasoned argument.

The result was read in AI-safety commentary as a partial explanation for why arguments citing single-digit extinction probabilities from AI have not translated into public demand for AI to be treated as an overriding policy priority, even among people who accept the underlying risk estimate.