Used Car Salesbots? Honesty and Credulity of LLMs as Bargaining Agents under Partial Information
arXiv:2605.31445v1 Announce Type: cross Abstract: In this work we study agents in simulated bargaining scenarios, where a buyer and a seller communicate through a text channel and attempt to negotiate mutually beneficial trades, under different information regimes (complete information, information asymmetry or mutual uncertainty). We evaluate their performance w.r.t. game-theoretical solutions and further investigate their honesty (their tendency to disclose or withhold information or to mislead and deceive) as well as their credulity (their tendency to trust or distrust information provided
The proliferation of LLMs makes their use as autonomous agents in complex, adversarial scenarios an immediate research and deployment priority.
Understanding the honesty and credulity of LLMs in bargaining scenarios is critical for developing robust, trustworthy AI agents capable of interacting with humans and other AI systems.
This research provides a framework for evaluating the strategic and ethical dimensions of LLM behavior in negotiation, moving beyond simple task completion to nuanced social interaction.
- · AI developers focused on agentic systems
- · Companies seeking to automate sales and negotiation
- · Researchers in game theory and AI ethics
- · Companies deploying insufficiently tested AI for high-stakes interactions
- · Human negotiators without augmented tools
LLMs will begin to be deployed in complex bargaining and sales roles, potentially enhancing efficiency.
The ethical implications of AI deceptiveness and gullibility in commercial transactions will necessitate new regulatory frameworks and transparency requirements.
A competitive landscape will emerge where AI agents are trained specifically to out-negotiate other AI agents, leading to an AI arms race in strategic interaction.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG