Artificial Intelligence and the Transformation of the Labor Market: A Markov Game Framework

Abstract: Abstract: [Objective] This study integrates Acemoglu & Restrepo's task-based model with the Stochastic Game framework to construct a computable four-state dynamic game system, aiming to analyze the interaction mechanism between artificial intelligence technological progress and labor market structure. [Methods] Taking the automation margin and human capital structure as state variables, the continuous task space is discretized into four stages: "traditional economy period—transitional pain period—human-machine collaboration period—post-AI steady state period." Through the joint specification of single-period payoff matrices, state transition matrices, and Bellman equations, the strategic interaction between firms and workers and the Markov Perfect Equilibrium are characterized. Furthermore, three policy instruments—training subsidies, automation taxes, and matching efficiency improvements—are introduced to analyze their structural impacts on payoff matrices and transition matrices. [Results] The impact of artificial intelligence on employment exhibits significant stage-dependent nonlinear characteristics, with multiple equilibria and path dependence: the system may converge to a high-efficiency equilibrium of human-machine collaboration, fall into a low-level trap due to skill mismatch, or collapse due to excessive automation. The substitution effect is most pronounced during the transitional pain period, while the combined force of productivity effects, automation deepening effects, and recovery effects determines the direction of state transitions. [Limitations] The discretization of state space simplifies the continuous technological evolution process; parameter calibration lacks micro-empirical data support; the two-player setting does not incorporate governments and financial institutions into the endogenous game framework. [Conclusions] Policy interventions require combined measures: "training subsidies + matching efficiency improvements" should be implemented during the transitional pain period, "automation taxes" should be introduced during the steady state period to prevent excessive automation, and inequality warning mechanisms based on wage ratios should be established.