AI-driven income inequality and preferences for redistribution

Theoretical and early empirical work in economics suggest that commercial deployment of advanced artificial intelligence (AI) can have significant implications for economic inequality. This paper provides the first empirical insights on how may people’s preferences for redistribution respond to such AI-induced inequalities. In a stated choice experiment, we let participants ( n = 750) choose their preferred income distributions in scenarios where initial income differences can vs. cannot be attributed to the use of AI technology. Preregistered data analyses indicate that participants’ preferences shift towards more equal income distributions in scenarios with AI. However, the shift is not dramatic and on its own does not point to a need for a fundamental restructuring of redistributive policies. In exploratory analyses we test whether participants’ previous experience with AI, their beliefs about AI’s future economic impacts, or their political preferences moderate the treatment effect.