The Fear of AI is the Fear of Looking in the Mirror: Low Trait Honesty-Humility Predicts Individual Perceived Likelihood of Future Apocalyptic AI Scenarios Through Beliefs in a Competitive World

Abstract

Apocalyptic narratives about Artificial Intelligence (AI), in which a future superintelligence dominates or annihilates humanity, have become a prominent source of public anxiety and a central concern in the AI alignment discourse. Because the behavior of a hypothetical superintelligent AI is fundamentally unknowable, individual expectations about such scenarios may reflect projections of users’ own social worldviews and self-perceptions rather than grounded technological forecasts. Drawing on Duckitt’s dual-process model of ideological attitudes, we tested whether Honesty-Humility, beliefs in a competitive world, and Social Dominance Orientation (SDO) predict the perceived likelihood of apocalyptic AI futures in a preregistered online survey of a representative UK sample (N = 300). Serial mediation analysis revealed that low trait Honesty-Humility indirectly predicted AI domination beliefs through competitive worldview, supporting a "projection" interpretation: individuals who perceive themselves as manipulative and exploitative expect the same of a hypothetical superintelligent AI. This effect is robust even after controlling for AI literacy, AI attitudes, AI use, and demographics. We discuss implications for understanding AI anthropomorphism, AI anxiety, and the socio-psychological roots of apocalyptic AI narratives.