Innovation & Culture
Humanitarian aid workers often enter conflict zones with limited conflict knowledge and minimal on-ground records, making it harder to negotiate access and reducing the impact of their efforts
Our goal was to efficiently equip aid workers with conflict insights, enable secure relationship tracking, and support institutional record-keeping for more effective negotiation
I led research, design, and testing for our web and mobile apps on this project with my thesis partner, and we built a working prototype of the products
Concordance was highly valued by humanitarian aid workers. We also collaborated with leading peace tech company Culture Pulse, who shared the UNDP's interest in our products
For my thesis, my partner and I developed a suite of AI tools designed to empower humanitarian aid workers with historical data, real-time insights, and relationship records to enhance frontline negotiations. Although we collaborated on all aspects of the project, my primary focus was on the research and design of the products.
Since 2004, international conflict has surged ninefold, making aid delivery a major challenge for organizations like MSF, Oxfam, and the IRC. Aid workers, known as frontline negotiators, must secure access to deliver life-saving supplies. When negotiations fail, they risk their lives, hinder aid delivery to thousands, and jeopardize their organization’s progress in the region.
These frontline workers negotiate with adversaries to facilitate the delivery of aid, education, sanitation, and other essential services in conflict zones. They also assist with camp coordination and shelter provision. However, negotiating in these volatile environments can be perilous; if negotiations fail, they not only jeopardize aid efforts for vulnerable populations and the success of their organizations, but in extreme cases, they also risk their own lives.
Over three months, we engaged over 40 experts in conflict resolution, humanitarian aid, and technology through ethnographic studies and interviews to understand the challenges negotiators face in securing aid access. During the course of our thesis, we also partnered with Culture Pulse, a Peace Tech company providing conflict-monitoring tools for the UN, who became our advisors and granted us access to their data and APIs. We also conducted rapid prototyping and co-creation sessions with frontline negotiators and design experts, leading to two key insights.
Because of lack of time negotiators often enter conflicts with little knowledge. This can risk aid delivery and the lives of the negotiators themselves
“[Frontline negotiators] stay in a country for a very short time, and enter high-stakes conflict without any knowledge of the background or culture”
— Paula Soumaya Domit Negotiator, MPP Candidate, HKS
“There was a female doctor with us in Iraq... she didn’t have enough knowledge of the local culture and assumed the camp was safe and went out for a run in shorts... people started to spit [at her] and it started to get violent”
— Pavol Kosnáč Director, DEKK Institute
Personal and institutional records are essential to develop relationships. Relationship management is vital for safety, yet, key information rarely gets recorded
“Developing relationships is vital for your own security. If people don’t trust you, and [you] don’t have relationships, you can get caught in a shelling or another dangerous situation.”
— Martha Thompson Professor, MIT, D-Lab
“We didn't have records [of past negotiations] - It was ok when we were small, but became a problem when our project grew. People who came after couldn’t learn because people left without any knowledge transfer”
— Pavol Kosnáč Director, DEKK Institute
We initially designed a conflict prediction tool to analyze news data and gauge on-ground sentiments within different factions of a community. By applying conflict modeling heuristics developed by Claude Bruderlein, our goal was to predict when tensions might escalate into conflict and intervene before violence occurred or lives were lost.
We tested our preliminary prototype with experts and quickly discovered key challenges. Many conflict zones lack readily available data, and in areas where data exists, its accuracy is often unreliable. This makes predictive tools potentially dangerous, as they may provide incomplete or misleading insights. Additionally, conflicts only become predictable after prolonged periods, when there is a substantial amount of accessible information, limiting the tool’s effectiveness in early stages of unrest.
“We don't collect any data on the field or records of on-ground information and interactions. Whatever people know is from their own experiences”
— Fuad Mallick, Professor and Dean, School of Architecture and Design at BRAC University, on fieldwork in the Rohingya camps
We also designed a tool for humanitarian aid organizations to simulate negotiations, enabling negotiators to practice before entering discussions. This tool could also be utilized for negotiation training for new aid workers.
Future Opportunity
During testing, we discovered that several organizations, including one at Harvard, are already developing similar conflict prediction tools. However, their accuracy is still limited by the availability and quality of data. This revealed an opportunity to build a tool that enables on-the-ground data capture, which could enhance institutional knowledge and improve future conflict negotiation tools. However, we also learned that frontline negotiators are reluctant to collect such data due to its potentially sensitive and incriminating nature. Therefore, any tool we create must prioritize security and give negotiators full control over what data they share and how much is disclosed.
We saw the gaps as opportunities to provide frontline negotiators with tools that bring them up to speed about a conflict and to help them maintain a personal and institution record of on-ground relationships and negotiations. This led us to our solution Concordance. The Concordance network consists of two key tools: Harbor and Bridge. Harbor works on the micro level, helping negotiators manage their relationship networks. Bridge works on the macro level, helping negotiators understand the conflicts they work in.
We created multiple sections focused on art education, featuring comprehensive artist profiles with CVs, notable acquisitions, video content documenting the artists’ practices, and voice notes for a more immersive learning experience. Additionally, we included profiles of galleries, an art history timeline, curated virtual exhibitions, and article sections, allowing users to explore artists and their work through rich, experiential content.
We learned that frontline negotiators often move between contexts quickly and lack the time to fully understand the history, factions, and sentiments of a conflict. Bridge helps negotiators get up to speed, especially when time is limited. Research shows that having specific, detailed information improves negotiation outcomes by allowing negotiators to cite clear examples of challenges and risks. Our tool provides contextualized knowledge, enabling users to negotiate more effectively and with greater precision.
We created multiple sections focused on art education, featuring comprehensive artist profiles with CVs, notable acquisitions, video content documenting the artists’ practices, and voice notes for a more immersive learning experience. Additionally, we included profiles of galleries, an art history timeline, curated virtual exhibitions, and article sections, allowing users to explore artists and their work through rich, experiential content.
People often store notes in Word docs or handwritten form and hesitate due to security concerns. Harbor simplifies this with secure, anonymized storage and user control over shared data. LLMs summarize conversations, helping users quickly catch up and access past knowledge, enabling stronger relationships and more effective negotiations, especially for newcomers in conflict zones.
Our backend operates through a straightforward data summarization process with four key steps: pulling relevant data, cleaning it, summarizing it using GPT-4, and serving it in our UI. We leverage CulturePulse’s API for pre-summarized data like news and sentiment analyses, while for other sources—such as historical info, PDFs, and proprietary data—we perform our own summarization. We also integrate expert advice from Harvard's IT department to ensure strong data privacy and security, following recommendations from their AI sandbox initiative.
“This is great! It would be a helpful starter pack for a new person coming to the environment, many organizations don’t provide that”
— Pavol Kosnáč Director, DEKK Institute
We presented our tools to frontline negotiators and received enthusiastic feedback. However, we learned that negotiators often focus on specific aid clusters, such as water infrastructure, meaning the information should be tailored to their specific needs. Additionally, different organizations require varying types of data, so the tool needs to be modular and adaptable. Despite the positive response, some negotiators expressed concerns about security, particularly if their phones fall into the wrong hands. This raised the question of whether further encryption methods can be implemented to ensure data safety.
— Claude Bruderlein Professor, Frontline Negotiation, HKS
— Martha Thompson Professor, MIT, D-Lab
Next Project
OBEROI HOTELS
Driving sales with a redesign that embodies the world's most luxurious hotel chain
SBI CARD