The Human Utility Factor: A Computable Welfare Metric That Reframes AI Governance as a Constrained Optimisation Problem
创建于 更新于
摘要
This paper formalizes a macro-socioeconomic stability constraint for AI governance and introduces the Human Utility Factor (HUF), a differentiable welfare metric decomposed into Agency, Wellbeing, and Economic Stability. It links HUF to actionable levers—automation depth ha, redistribution α+β, and employment coverage ρ—and derives a closed-form interior optimum h∗a and a redistribution threshold (α+β)∗. Through three-agent Stackelberg MARL simulations across BAU, Partial, and Nordic regimes, the study shows redistribution is the decisive lever and that Wellbeing-dominant equilibria can arise without a floor on redistribution, highlighting governance risks. Empirical evidence (GAGI divergence, AI layoffs, labor share decline) supports the urgency of a system-level stability constraint. The framework recasts AI governance as a computable constrained optimization problem with enforceable inequalities that regulators can monitor and enforce. [page::0][page::4]
