As climate change, conflict, and political instability continue to displace millions, humanitarian agencies face a vital question: can artificial intelligence help predict these crises before they unfold?
By Stuart Kerr, Technology Correspondent
Published: 20 July 2025
Last Updated: 20 July 2025
Contact: liveaiwire@gmail.com | Twitter: @LiveAIWire
Author Bio: About Stuart Kerr
Predicting the Unpredictable
In recent years, machine learning has begun transforming how aid organisations respond to forced displacement. Traditionally, responses were reactive—relief only arrived after borders were breached. Today, however, AI offers a proactive alternative.
The World Bank has piloted projects that use 90+ variables to forecast refugee flows 4 to 6 months in advance. By analysing factors such as food insecurity, weather patterns, and political instability, these algorithms are helping governments prepare aid stations, shelter logistics, and security responses before a crisis hits.
From Model to Movement
This predictive approach is not theoretical. Stanford's GeoMatch model uses machine learning to recommend optimal resettlement destinations for refugees based on long-term employment outcomes. When tested, it significantly improved refugee success rates.
Closer to home, the Danish Refugee Council is using forecasting models to anticipate forced displacement 1 to 3 years ahead. These tools inform strategic planning, not just emergency relief. Predictive models are also detailed in the World Bank PDF overview.
The Promise and the Pitfalls
While AI adds foresight, it also brings risk. Your article The Algorithm Will See You Now shows how quietly decision-making authority has shifted to machines. Applying that same logic to refugee status, border policy, or relocation could remove critical human judgment from life-altering decisions.
In Raising Children with Code, you noted that predictive AI can inadvertently shape the very futures it attempts to foresee. In displacement contexts, this could mean reinforcing stereotypes, misclassifying risk factors, or privileging data-rich populations over those left off the digital map.
There’s also the question of consent. Refugees rarely opt into these systems. The models rely on passive data collection from satellites, border scans, or aid distribution patterns. As highlighted in AI and the Shadow Economy, this can lead to profiling or surveillance practices that go unchecked.
Guardrails and Governance
To guide ethical implementation, the Chatham House report Refugee Protection in the Artificial Intelligence Era urges policymakers to preserve human oversight and legal accountability. It warns of AI being used to automate exclusion or fast-track asylum decisions without due process.
The future depends on balance. As the UNHCR & UNDP joint report suggests, AI should serve inclusion, not control. Used responsibly, it can improve access to services and match people to safer, more productive environments.
Anticipation, Not Automation
Forecasting displacement is not the same as determining fate. Algorithms must support, not replace, the judgment of humanitarian actors. Predictive models are a tool—powerful, but not omniscient.
If we treat their output as prophecy rather than guidance, we risk turning crisis prediction into crisis production.
About the Author
Stuart Kerr is the Technology Correspondent for LiveAIWire. He writes about artificial intelligence, ethics, and how technology is reshaping everyday life. Read more