Deciphering Multilateral Climate Negotiations with Language Models
Keywords: negotiations, climate, multilateral, llm, policy
Abstract: Multilateral negotiations are crucial for effective climate actions. These negotiations follow complex, multi-step procedures involving representatives from 198 different countries. Negotiators representing the interests of financially constrained countries are at a serious disadvantage: (i) they face language barriers, (ii) have limited experience, and (iii) operate in smaller teams. In this work, we outline several ways in which large language models (LLMs) can alleviate these hurdles. We formalize the negotiation problem using recent advances in LLM agency and propose several modules based on interviews with climate youth negotiators conducted at COP28 in Dubai. We argue that LLMs could represent a “chess moment” for negotiations and hope our work can convince more NLP researchers to contribute to climate negotiation research.
Archival Submission: arxival
Submission Number: 33
Loading