Page 449 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

and reliability of RiMAP is dependent on the ability to continuously update its content in a
timely and efficient manner.

Objectives and aims of the use case 4.4-Productivity

This use case forms part of the broader development of the Rights Mapping and Analysis
Platform (RiMAP). As part of the ongoing efforts to enhance RiMAP, UNHCR is integrating the
internal Virtual Legal Assistant (VLA) powered by Large Language Models (LLMs) to facilitate
legal data collection, improve policy analysis, and make this information more accessible to
internal users, focusing on data translation, extraction summarization as well as production of
structured outputs.

Primary Objective (Internal Focus): To support UNHCR country editors and legal advisors by
developing and integrating an internal Virtual Legal Assistant (VLA) powered by LLMs. The
VLA will process and analyze large volumes of legal documents from national sources across
all UN countries/territories, facilitating legal data collection (extraction, summarization) and
improving policy analysis efficiency. Target: Aim to reduce preliminary legal research time for
targeted tasks by 30% within 12 months of full deployment.

Once data is collected, the AI assistant is able to extract key legal insights and summarize complex
legal texts. This includes government publications, legal databases, academic research, and
other publicly available resources related to the legal rights and status of displaced and stateless
persons. By leveraging AI, the VLA will enable UNHCR to analyze evolving legal landscapes,
ensuring that the organization stays up to date with legal developments. By automating this
process, the VLA will greatly reduce the manual effort required to track and gather legal data,
ensuring that the platform remains current without the heavy workload traditionally involved
in such research.

The secondary goal of this use case is to expand the VLA’s functionality to support users. By
transforming the VLA into an AI-powered chatbot, the system provides easily accessible and
user-friendly legal information derived from the RiMAP library. This will be a significant step
towards making legal data more accessible within UNHCR but importantly also to a wide range of
stakeholders, including universities, legal aid organizations, governments, development actors,
civil society groups, and, importantly, forcibly displaced and stateless persons themselves.

The anthropomorphic chatbot interface offers an intuitive interaction that allows users to
interact with the VLA with natural language, ask questions, and receive legal information in
a more conversational and accessible format. The VLA’s capabilities will span across all UN
member states and territories, ensuring that users have access to a comprehensive, global legal
resource. This will empower users to navigate complex legal systems and better understand
the rights of displaced and stateless persons, enabling more informed advocacy and facilitating
greater legal support for such populations.

Innovative technological approach implemented

The VLA automatically scrapes, extracts, and categorizes relevant legal documents from these
diverse sources, allowing it to build and expand its repository of up-to-date legal information
continually. The VLA automates the extraction and summarization of relevant legal data from
vast volumes of legal text corpora, policy documents, and public legal databases.

413