Building an AI Chatbot for Refugee Legal Information
Lead Researcher at Refugee Solidarity Network
Skills
Technologies
Team Size
2 people
I designed and built a proof-of-concept LLM-powered chatbot to help asylum-seekers access legal information about their rights in Türkiye, then published our findings to help other humanitarian organizations navigate AI adoption responsibly.
Context
In 2023-2024, Refugee Solidarity Network recognized the transformative potential of AI and sought to explore how these technologies could enhance access to justice for asylum-seekers. Rather than theorize, we decided to build something real: a chatbot that could answer questions about refugee rights, grounded in authoritative legal sources.
What I Built
- 1
Development Environment
Set up a collaborative workspace using Gitpod with Python and transient Docker containers, enabling rapid onboarding and team-based development.
- 2
Data Pipeline
Built pipelines to curate and transform multilingual legal information from the Refugee Information Portal across seven languages, with transparent source citation.
- 3
RAG Chatbot
Implemented a Retrieval Augmented Generation system using LangChain, ensuring responses were accurate, contextually relevant, and grounded in the curated knowledge base.
- 4
Evaluation & Report
Conducted comprehensive evaluation of response quality and documented findings, including ethical considerations and limitations for the humanitarian sector.
Impact
The published report was downloaded over 600 times and was recognized by the UK Humanitarian Innovation Hub, which included it in their Directory of AI-Enabled Humanitarian Projects as an example of responsible AI innovation.
Navigating Humanitarian AI: Lessons Learned from Building a Chatbot Proof-of-Concept
Full report published on ReliefWeb
reliefweb.int