LLM Agent Emotions Significantly Impact Negotiation Outcomes

This study investigates the impact of emotions on large language model (LLM) agents during price negotiations. Researchers assigned various emotional states, such as anger, happiness, and fear, to buyer and seller agents, who then negotiated prices for real consumer products. The findings indicate that emotions profoundly affect negotiation dynamics and outcomes. For instance, angry buyers rarely reached agreements, while happy buyers agreed more often but secured less favorable prices than fearful buyers. The study also highlighted that the role of the agent (buyer vs. seller) influenced how emotions manifested, with buyer emotions primarily driving acceptance or rejection, and seller emotions affecting concession patterns. These emotional influences extended beyond mere language to impact negotiation termination and price trajectories, raising important considerations for the design of emotion-conditioned AI agents in commercial applications.