Negotiation Backtable Bots: Using GenAI to Improve Multiparty Negotiation Instruction

Abstract The authors developed negotiation coaching bots using generative artificial intelligence (GenAI) and experimented with them in multiparty negotiation courses at MIT in the spring of 2024. This article focuses on one of the three kinds of bots we created: a coaching bot that allows learners to interact with the back tables of their assigned negotiation partners in an upcoming multiparty role-play simulation. Interacting with the back tables of other parties before entering into human-human negotiations can improve negotiation preparation in important ways. Interacting with backtable bots made very clear to students that negotiation parties are non-monolithic entities. We include descriptions of all the learning steps that our students went through, and quote them describing their reactions to interacting with a backtable bot and being coached by a GenAI negotiation coaching bot. We summarize our seven steps to building GenAI negotiation bots, which are applicable to all three types: preparation coaching bots, debriefing coaching bots, and backtable bots. We provide examples from our work on the backtable bots, and pay specific attention to the importance of prompt design. We conclude with recommendations for further development of negotiation coaching bots.