Page 290 - United Nations Activities on Artificial Intelligence (AI) 2024

United Nations Activities on Artificial Intelligence (AI)

UNHCR's proGres registration data with high-resolution satellite imagery and advanced
spatial modeling techniques. This innovative combination aimed to create more precise
sampling frames, enabling the identification of refugee populations in out-of-camp
settings. By leveraging data science and AI-driven analysis, the team sought to overcome
the limitations of outdated or non-specific sampling frames, ensuring that surveys could
be conducted more efficiently and effectively.
The implementation of this solution has led to significant improvements in the accuracy
and efficiency of refugee surveys. By utilizing AI and data science, UNHCR can now better
understand the living conditions and needs of displaced populations, facilitating more
targeted and effective humanitarian responses. This approach not only optimizes resource
allocation but also ensures that the voices of refugees are more accurately represented
in data collection efforts.
The impact of this project extends beyond immediate survey improvements. It sets a
precedent for the integration of AI and data science in humanitarian efforts, demonstrating
how innovative technologies can address complex challenges in refugee data collection.
By refining survey methodologies, UNHCR enhances its capacity to respond to the
evolving needs of displaced populations, ultimately contributing to more effective and
responsive humanitarian interventions.
• Department/Division: UNHCR Innovation Service, Data Innovation Fund 2023-2024,
Data, Identity Management and Analysis (DIMA) units, UNHCR Global Data Service (GDS)
• Project Type/Output: AI model
• Project Status: Completed
• Project Start Year: 2023
• Project Domain: Forced displacement, Human rights, statistical analysis, camp
management
• Data Source: PRIMES/ProGres UNHCR registration data
• Data publicly available: No
• Technology/Platform: open-source/R
• Related Sustainable Development Goals (SDGs): SDG 10 – Reduced Inequality; SDG 16
– Peace, Justice, and Strong Institutions; SDG 17 – Partnership for the Goals
• Partnership(s)/Collaborator(s): University of Southampton, WorldPop
• Relevant Links and Multimedia:
https:// jhumanitarianaction .springeropen .com/ articles/ 10 .1186/ s41018 -024 -00157 -6

• Lesson Learned: Implementing AI-driven approaches in humanitarian contexts has
yielded valuable insights. A key lesson is the importance of integrating diverse data
sources—such as registration databases, satellite imagery, and spatial covariates—to
enhance the precision of refugee population mapping. For instance, in Cameroon,
combining these datasets enabled the disaggregation of refugee populations into
100-meter grid cells, providing high-resolution spatial information crucial for targeted
interventions. This integration not only improved the accuracy of population estimates
but also informed more effective service delivery and resource allocation strategies.
However, the deployment of AI in humanitarian settings presents several challenges.
One significant difficulty is ensuring data privacy and security, especially when handling
sensitive information of vulnerable populations. Additionally, the complexity of AI models
can lead to a lack of transparency, making it challenging for stakeholders to understand
and trust the outcomes. There are also concerns about algorithmic bias, which can
inadvertently perpetuate existing inequalities if not properly addressed. Furthermore,
resource constraints, including limited technical infrastructure and expertise, can hinder
the effective implementation of AI solutions. These challenges underscore the need for
robust ethical frameworks, stakeholder engagement, and capacity-building initiatives to
ensure responsible and effective use of AI in humanitarian efforts.

272