Peace on Earth

Peace on Earth is an AI-powered conflict resolution platform that helps opposing factions find common ground through intelligent, confidential mediation.

Comment

Inspiration

Can we use AI for something really good and practical? While everyone was generating images and writing code, we asked: what if AI could mediate real conflicts? From water disputes between nations to trade wars threatening global stability—countless conflicts persist because finding neutral ground is hard. We built Peace on Earth to prove AI can bridge divides when human mediators are unavailable, biased, or distrusted.

What it does

Peace on Earth is a browser-based AI conflict resolution platform. An admin creates a mediation session between two opposing factions (e.g., nations, labor unions, political groups). Each faction submits their public demands, grievances, and—critically—private interests and willing concessions through independent, secure links. Once both sides submit, Google's Gemini AI analyzes the combined data and generates a balanced peace proposal. The system protects confidentiality (private interests are never shared directly) while still using that information to craft workable solutions. Factions review proposals and indicate acceptance or rejection.

How we built it

  • Framework: CleanLit (AI-First Frontend Framework) with design system for modern glassy styling
  • AI Integration: Google Gemini 3 for proposal generation and mediation logic
  • Architecture: Client-only implementation using LocalStorage for state management—no backend server required for POC
  • Security: Distributed links ensure factions cannot see each other's submissions; only the AI sees combined data
  • Deployment: Static site deployed to Netlify for instant accessibility

Challenges we ran into

  1. Confidentiality vs. Utility: How to make the AI generate meaningful proposals without exposing sensitive private interests to either faction. Solved by clearly separating public and private data fields in prompts.
  2. Prompt Engineering: Getting Gemini to act as a truly neutral mediator—not favoring one side, not being too vague, and proposing concrete, actionable solutions.
  3. State Synchronization: Without a backend, coordinating state between admin dashboard and faction views required careful LocalStorage design.
  4. Realism: Balancing simplified demo scenarios with enough complexity to showcase the platform's potential for genuine conflicts.

Accomplishments that we're proud of

  • Created a working end-to-end conflict mediation flow in a single hackathon
  • Designed a system where opposing parties can negotiate without direct contact—critical for high-stakes disputes
  • Successfully handled realistic scenarios that have stumped opposing parties for years
  • Zero backend required (for POC only)—making it deployable anywhere, instantly
  • Clean, intuitive UI that works for both tech-savvy and non-technical users

What we learned

  • AI neutrality requires careful prompting: Without explicit instructions, LLMs can subtly favor the first or last input in their context window.
  • Private interests drive solutions: Public demands often deadlock; knowing what's truly negotiable (private concessions) unlocks breakthroughs.
  • Simplicity enables adoption: Removing complexity means NGOs, governments, or even families could use this without IT support.
  • Conflict has patterns: Water rights, trade, territory—different domains share structural similarities that AI can recognize and leverage.

What's next for Peace on Earth

  • Multi-party support: Expand beyond two-faction conflicts
  • Persistent backend option: For organizations needing audit trails and data persistence Integration with diplomatic channels: APIs to export proposals to UN, regional organizations, or government systems
  • Historical case learning: Train on past successful mediations to improve proposal quality
  • Mobile app: Enable field negotiators and diplomats to mediate anywhere

Built With

Share this project:

Updates