About UNHCR
UNHCR, the UN Refugee Agency, is a global organization dedicated to saving lives, protecting rights and building a better future for refugees, forcibly displaced communities and stateless people.Activities by UNHCR
UNHCR Responsible AI Framework and Governance Initiative
Global DenmarkThis initiative establishes a risk-based, responsible AI framework for UNHCR to guide ethical design, procurement, deployment, and use of AI systems. It operationalizes UN AI principles through governance, risk classification, DPIAs, and monitoring, with future plans to include tools.
Activity Type Technical AssistanceResearch/Reports/AssessmentsPolicy/Regulatory GuidanceRiMAP VLA: A Virtual Legal Assistant for Scaling Humanitarian Law and Policy Analysis on the UNHCR Rights Mapping and Analysis Tool (RIMAP)
Global Not Applicable
UNHCR developed a Virtual Legal Assistant (VLA) powered by Retrieval Augmented Generation (RAG) into its Rights Mapping and Analysis Platform (RiMAP). This platform support UNHCR country editors by efficiently collecting, processing, and analyzing vast amounts of legal documents across all UN countries and territories. This enables legal research and analysis processes, including extraction, translation, summarization, and drafting.
Once data is manually collected and indexed in a library accessible to the AI assistant, a first draft of responses to questions can be generated. This draft is created by extracting relevant paragraphs, translating them, and summarizing the content to produce an accurate response with sources. This draft serves as the foundation for the legal analysis that country editors can verify and further edit.
The library of sources includes government publications, laws and regulations, policies, academic research, and UNHCR reports. By leveraging AI, the VLA allows UNHCR to significantly reduce the time spent on legal research and drafting, ensuring the timely and comprehensive completion of RiMAP within a reduced timeframe.
The VLA is an AI-powered chatbot that utilizes resources contained within RiMAP. This public-facing chatbot provides easily accessible and user-friendly information on the rights of forcibly displaced and stateless persons. This initiative marks a significant step towards making legal data more accessible to a diverse range of stakeholders, including universities, legal aid organizations, governments, development actors, civil society groups, and, most importantly, forcibly displaced and stateless persons themselves.
Understanding population movement related to COVID-19 border closures
Latin America and the Caribbean BrazilVenezuelaThis project aims to calculate and anticipate the number of displaced persons crossing the Brazil-Venezuela border to understand their need for humanitarian support. It consists of a queue modeling tool, a nowcasting effort using big data, and predictive models for forecasting future movements.
Activity Type AI Tools/SolutionsResearch/Reports/AssessmentsProject Jetson
Africa Somalia
Project Jetson (2017-2021) was a predictive analytics experiment to discover, understand, and measure the specific push and pull factors that cause, indicate or exacerbate the forced displacement. Project Jetson gives UNHCR and other humanitarian organizations the potential to become more proactive in their contingency planning, emergency preparedness and response efforts, a transformation that could significantly improve on-the-ground relief services and more importantly the lives of those who are forcibly displaced.
Humanitarian challenge Drought exacerbating internal displacement in Somalia generated by protracted conflict.
Challenge Question How might UNHCR predict the number of people arriving into the different regions of Somalia?
The variables selected in Jetson, represent some of the most important influencing factors for forced displacement in the operational context of Somalia. The variables were selected according to different assumptions we needed to later validate, even from qualitative data or quantitative ground truth data. For selecting variables and validating them, we used several innovation methodologies,