Designing AI for Humanitarian Frontline Negotiations

A team of researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) is exploring how AI tools like ChatGPT could support some of the world’s most delicate conversations: humanitarian frontline negotiations. In a paper published in the Proceedings of the 4th Annual Symposium on Human-Computer Interaction for Work, a team led by researchers from the Center for Human-driven AI Research and Methods at SEAS and negotiators from Frontline Associates, worked with 32 experienced negotiators to understand where AI might help — or hinder — their work. 

Rather than asking AI to choose tactics or speak on their behalf, participants said they needed help making sense of complex situations: organizing scattered information, mapping stakeholders, exploring possible compromises, and assessing risks. Building on these insights, the team designed a prototype interface in which an AI system guides negotiators through structured steps and risk assessment matrices. The interface helps negotiators maintain full decision-making authority while helping them to visualize negotiation scenarios, identify red lines and bottom lines, and explore possible agreements.

Negotiators preferred this “process support” role and were more comfortable using AI when it clearly enhanced, rather than replaced, their professional judgment and relationships with counterparts. The study highlights design principles for deploying AI in other high‑stakes fields — from diplomacy and crisis response to healthcare and social services — where context, ethics, and trust are as important as outcomes. The researchers argue that in these fields, AI should enhance human judgment and adaptability rather than automate complex decisions requiring cultural sensitivity and ethical reasoning.

Read the full paper here.