AI + Negotiation

Since 2023, the Data to Actionable Knowledge (DtAK) lab has been leading an interdiciplinary research initiative to understand the impact of AI on frontline humanitarian negotiation.

Humanitarian negotiations in conflict zones, known as frontline negotiation, are often adversarial, complex, and high-risk. Negotiators must navigate a wide range of conflicting perspectives by accurately and quickly synthesizing large amounts of unstructured information from various sources. Given recent advances in the ability of large language models (LLMs) to synthesize and extract patterns from natural language data, communities of negotiation practioners have begun to experiment with using LLM-based AI-tools to support their workflows.

However, since frontline negotation is high-stakes and risk-sensitive, it is important for us to systematically study how AI tools can be responsibly and ethically developed and used in these contexts. Our research focuses on three areas:

1. Understanding Human+AI Interactions in Frontline Negotiation: we aim to understand where (and how) best AI can support the decision processes in negotiation. Specifically, we study the interactions between human practitioners and AI decision-support systems for different negotiation tasks, as well as the ways AI model properties (e.g. limitations of LLMs) can impact human users.
2. Leveraging Modern AI for Situational Awareness: we aim to evaluate the capabilities of LLMs when used for situational awareness in frontline negotiation contexts. Specifically, we study AI+human frameworks that improve the quality of information extraction/synthesize provided by LLMs by leveraging human domain-knowledge.
3. Exploring Complementarity in Combining Traditional and Modern AI Approaches: we aim to understand how to combine traditional machine learning approaches with modern AI for reasoning and planning tasks in frontline negotiation. We aim to improve task-performance as well as the steerability and interpretability of AI systems in these contexts.

Core DtAK Members

A number of DtAK members are actively working in the AI+Negotiation research initiative:

• Weiwei Pan
• Finale Doshi-Velez

Research Collaborations

We work closely with researchers in Human-Computer Interaction (HCI) and practitioners from the frontline humanitarian negotiation community. Our collaborators include:

• Krzysztof Gajos and Zilin Ma of the Intelligent Interactive Systems Group, at the Harvard School of Engineering and Applied Sciences
• Claude Bruderlein at the Harvard Kennedy School of Government
• Frontline negotiation practitioners, through Frontline Associates
• The Computational Science and Engineering Capstone Course at the Harvard School of Engineering and Applied Sciences

We are always interested in new research connections! For collaboration, please reach out to Weiwei Pan (weiweipan at harvard dot edu).

Research Support

In 2025-2026, our work will be partially supported by a research grant from Cooperative AI.

Publications

“ChatGPT, Don’t Tell Me What to Do”: Designing AI for Context Analysis in Humanitarian Frontline Negotiations, Z Ma, Y Mei, C Bruderlein, K Gajos, W Pan

Activities and Events

• In March 2025, DtAK, in collaboration with Frontline Associates, will be co-hosting an AI+Negotation Hackathon (with a focus on climate negotiation) at Harvard University.
• During Summer 2024, DtAK, in collaboration with Frontline Associates, co-organized an online summer program on AI+Negotiation for practioners from different domains of negotiation (including climate negotiation, humanitarian access and migration).