Socioeconomic and digital divides in public support for AI-driven welfare decisions: cross-national evidence from 21 countries

As governments increasingly deploy artificial intelligence (AI) for automated decision-making, little is known about why citizens differ in their attitudes toward AI-driven decision-making in the public sector, especially in welfare governance, where decisions carry profound consequences for marginalized groups. As prior scholarship suggests that automated systems can act as a digital poorhouse that punishes low-income individuals, this paper examines whether there exists a socioeconomic divergence in public opinion and how a digital divide perspective helps explain this divergence across countries. Through a cross-national survey, the study finds that individuals experiencing economic hardship and holding negative attitude toward AI express significantly less support for AI-driven welfare decisions than their counterparts. The study further shows that the country-level divide in Human Development Index and internet penetration can further condition the effects of economic hardship and general AI attitude on support for AI-driven welfare decisions. Overall, the study not only highlights the importance of addressing the inequalities embedded in the politics of automation but also calls for future research to examine the role of the digital divide in shaping public opinion across policy domains and societal contexts.