AI, Political Opinion, and Behavior

Abstract

Technological innovation, from agricultural mechanization to computerization, has transformed economies in the past. Artificial intelligence (AI) represents the next wave of workplace automation, yet its effects on political preferences and voting behavior have remained unstudied so far. With AI predicted to affect most modern jobs, understanding who will be affected and how this influences political behavior is crucial. I develop an economic framework predicting AI automation winners (high-skill, high AI exposure) and losers (low-skill, high AI exposure). I argue that losers will favor more redistribution and oppose immigration, with the former mediating populist left voting and the latter mediating populist right voting. Using ESS data (2015-2024) with matching and mediation analysis, I find AI losers indeed favor more redistribution and less immigration, but only immigration attitudes translate into voting behavior, specifically for populist right parties. Notably, anti-immigration attitudes also predict populist left voting more than redistribution preferences. All in all, results demonstrate that (AI) automation losers misattribute economic hardship, changing political preferences and behavior in response to AI’s negative effects. This cautions against an AI transformation without policies supporting AI automation losers.