Personality Engineering: Experimental Evidence from Human-AI Negotiations

Abstract: Given the advanced capabilities of large language models (LLMs), many of the world’s largest firms have adopted LLM-based chatbots to negotiate with their customers and suppliers. For example, the multinational superstore Walmart, the telecommunications provider Vodafone, and the logistics company Maersk have all implemented LLM-powered chatbots to negotiate and maintain deals with their suppliers. This shift raises a question of immediate practical interest: How do negotiation outcomes with humans compare to those of chatbots? By incorporating customized chatbots into an integrative experimental design, we investigate this question while addressing a broader theoretical issue: How can agents—whether human or chatbots—balance warmth and dominance to optimize negotiation outcomes?

Keywords: Negotiation, LLMs, Human-AI Interaction