Improved community-based protection monitoring with AI and data science
Challenge
UNHCR Afghanistan gathers a vast amount of text-based qualitative data through community-based protection monitoring, but manually processing this information is time and energy intensive, resulting in lost opportunities to use this data to inform timely interventions.
Solution
Use AI-based software, particularly with natural language processing features to facilitate qualitative analysis of data gathered through community-based protection monitoring. This software would detect key words and phrases, categorise and translate them, and produce meaningful analysis.
Impact
Effective and efficient data analysis, resulting in better use of data to inform effective protection interventions and timely responses to protection risks
Project impact
Other information
In Afghanistan's shrinking protection space, UNHCR collects large volumes of qualitative protection data through community-based monitoring, border monitoring and focus group discussions, but capacity to analyse it manually is limited, leaving critical trends and advocacy opportunities unused. This project piloted, for the first time, the use of AI-powered qualitative analysis with NetBase Quid to automatically cluster narratives, detect sentiment, and surface protection trends from thousands of records in local languages. The project also produced the first UNHCR Data Protection Impact Assessment for AI-driven qualitative analysis and a reusable Standard Operating Procedure for anonymising qualitative datasets before upload, setting a precedent for safe, ethical use of AI in humanitarian protection work that can be replicated across other operations. The SOP is now ready and available for use by other entities in UNHCR.