Microeconomic Foundations of Multi-Agent Learning

TL;DR

This paper establishes an economic framework for multi-agent learning in markets, proposing a two-phase incentive mechanism that aligns individual learning with social welfare, supported by theoretical analysis and simulations.

Contribution

It introduces a novel economic foundation for multi-agent learning with strategic externalities and designs an incentive mechanism to optimize long-term social welfare.

Findings

Mechanism achieves sublinear social-welfare regret.

Coarse incentives can correct inefficient learning.

Incentive-aware design is crucial for safe AI deployment.

Abstract

Modern AI systems increasingly operate inside markets and institutions where data, behavior, and incentives are endogenous. This paper develops an economic foundation for multi-agent learning by studying a principal-agent interaction in a Markov decision process with strategic externalities, where both the principal and the agent learn over time. We propose a two-phase incentive mechanism that first estimates implementable transfers and then uses them to steer long-run dynamics; under mild regret-based rationality and exploration conditions, the mechanism achieves sublinear social-welfare regret and thus asymptotically optimal welfare. Simulations illustrate how even coarse incentives can correct inefficient learning under stateful externalities, highlighting the necessity of incentive-aware design for safe and welfare-aligned AI in markets and insurance.