Strategic Self-Improvement for Competitive Agents in AI Labour Markets

ICLR 2026 Conference Submission19996 Authors

Keywords: multi-agents, simulation, economics, self-improving, llm agents

TL;DR: We developed a simulated gig economy to study impact of AI to labour markets and identified strategic reasoning capabilities (metacognition, competitive awareness, long-term planning) that organically develop in successful agents

Abstract: As artificial intelligence (AI) agents are deployed across economic domains, understanding their strategic behavior and market-level impact becomes critical. We investigate the dynamics of an AI labor market through a simulated gig economy where agents controlled by fixed policies or Large Language Models (LLMs) compete for jobs, develop skills, and adapt their strategies under competitive pressure. Our analysis identifies three core capabilities that successful LLM-agents develop organically: \textbf{metacognition} (accurate self-assessment of skills), \textbf{competitive awareness} (modeling rivals and market dynamics), and \textbf{long-horizon strategic planning}. Moreover, we show that LLM agents explicitly prompted with these reasoning capabilities learn to strategically self-improve and demonstrate superior adaptability to changing market conditions. At the market level, our simulation reproduces classic macroeconomic phenomena found in human labor markets, while controlled experiments reveal potential AI-driven economic trends, such as rapid monopolization and systemic price deflation. This work provides a foundation to further explore the economic properties of AI-driven labour markets, and a conceptual framework to study the strategic reasoning capabilities in agents competing in the emerging economy.

Primary Area: transfer learning, meta learning, and lifelong learning

Submission Number: 19996