Renegotiable Options Fail to Prevent Hold-Up Between Two Profit-Maximizing Frontier LLM Agents
What it is
This paper investigates whether contract terms that prevent hold-up in human economic games also work for LLM agents. It simulates a hold-up scenario using Claude Opus 5.5 and GPT-6.1 Sol, finding that while humans invested more with renegotiable options, the LLM agents behaved according to strict game-theoretic predictions, failing to adapt when renegotiation was possible.
The problem it solves
Uncertainty about whether economic mechanisms designed for humans (like renegotiable options) effectively prevent hold-up problems when the parties are profit-maximizing AI agents.
What you could use it for
- Evaluating the economic rationality and strategic behavior of LLM agents
- Testing the applicability of human behavioral economics theories to AI agents
- Designing robust contract mechanisms for autonomous AI-to-AI transactions
Why it matters
It demonstrates that current frontier LLMs do not replicate human behavioral anomalies in economic games, suggesting that standard contract designs may fail to protect investment in AI-agent-mediated markets.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.