Page 451 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

reduce inequality, promote justice, and ensure the protection of displaced persons’ rights.
The use case will contribute to achieving SDG Goals 10 (Reduced Inequality) and 16 (Peace,
Justice, and Strong Institutions), ultimately empowering displaced populations by improving
their access to legal rights and protections. 4.4-Productivity

2�3 Future Work

The immediate next step is the development of an internal VLA prototype to support editors,
following which the VLA will be expanded to serve as a full-scale content generating tool,
focusing on the core functionalities, such as summarizing and researching large legal
documents, extracting legal data, and providing accurate legal analysis. The VLA will be tested
internally by UNHCR staff to validate the AI-driven features and to ensure the system can meet
the practical needs of legal research and policy analysis. Feedback from this phase will inform
the refinement and scaling of the VLA.

One of the key future goals is to fully automate the legal search and regular updates. Currently,
legal research often requires significant substantive manual input and time. In future iterations,
the VLA will leverage AI to automatically update RiMAP with relevant information pulled from
legal sites, making the process faster and more efficient. Automation will extend to continuous
data collection and updates, ensuring that the information in RiMAP is always current and
relevant.

Future enhancement will be the design of a public AI-powered chatbot for external users,
including universities, legal aid organizations, governments, civil society, and forcibly displaced
and stateless persons. This chatbot will serve as an intuitive interface, allowing users to ask
legal questions, search for relevant information, and receive clear, summarized legal guidance
directly from RiMAP’s library. The chatbot will be designed for ease of use, enabling external
stakeholders with varying levels of legal knowledge to access relevant and up-to-date legal
information quickly and efficiently.

Finally, future work will focus on extending language support to ensure broader accessibility
to legal information across regions. This will include more languages with specific legal
terminologies relevant to various geographical areas, enabling the VLA to cater to a wider
and more diverse user base.

3 Use Case Requirements

• REQ-01: Functional: The VLA shall summarize legal documents (up to 100 pages) into
5-15 key paragraphs with verifiable citations to the source document.
• REQ-02: Functional: The VLA shall extract predefined data points from designated legal
text types.
• REQ-03: Functional: The VLA chatbot shall respond to user queries using only verified
information in the RiMAP database, providing source links.
• REQ-04: Functional: The system shall automatically trigger a document's reindexing upon
detection of an update in its source or upon manual upload of a newer version.
• REQ-05: Non-Functional: Summarization requests shall be processed with an average
response time under 60 seconds.
• REQ-06: Non-Functional: The system must maintain an audit trail of AI suggestions and
human review actions.

415