What are the risks of using AI to make decisions about who receives aid

Genuine, High-Stakes Risks Requiring Careful Management

Using AI to make decisions about who receives humanitarian aid carries genuine, well-recognized risks, given how consequential these decisions can be for vulnerable populations, which is why responsible humanitarian organizations generally treat this as an area requiring particularly careful human oversight rather than full automation.

The Risk of Biased or Incomplete Data Producing Unfair Outcomes

If the data used to train or inform an AI-assisted aid allocation system reflects incomplete information about certain populations — perhaps because some groups are harder to reach for data collection, or because historical need assessment processes themselves contained bias — this can lead to unfair allocation decisions that systematically disadvantage specific groups, similar to documented bias risks that have emerged in other AI application areas.

The Risk of Reduced Human Judgment for Consequential Decisions

Aid allocation decisions often involve significant nuance and context — understanding a specific community’s particular circumstances, recognizing atypical situations that don’t fit standard assessment criteria, and exercising judgment about genuinely difficult tradeoffs when resources are insufficient to meet all documented need — a kind of contextual judgment that’s genuinely difficult for AI systems to fully replicate, creating real risk if human judgment is excessively reduced in favor of automated determination.

The Risk of Accountability Gaps When Decisions Go Wrong

When an AI-assisted aid allocation decision causes harm — wrongly excluding someone who genuinely needed assistance, for example — it can become genuinely unclear where responsibility lies: with the humanitarian organization deploying the tool, the technology developers who built it, or the specific data used to inform the decision, creating accountability gaps that are harder to resolve than when a decision was made directly and solely by an identifiable human decision-maker.

Why These Risks Matter More in Humanitarian Contexts Than Many Other Applications

These risks carry particularly serious weight in humanitarian aid contexts because the populations affected are often already highly vulnerable, the consequences of an incorrect decision can be severe (including genuine risk to health or survival), and affected individuals often have limited practical ability to challenge or appeal an aid allocation decision compared to, for example, a consumer disputing a commercial transaction.

Why Responsible Organizations Generally Maintain Meaningful Human Oversight

Given these genuine, serious risks, responsible humanitarian organizations generally maintain meaningful human oversight and judgment in aid allocation decisions, using AI-based analysis to support and inform this human decision-making rather than fully automating determinations about who does and doesn’t receive assistance.

Bottom Line

Using AI to decide who receives humanitarian aid carries genuine risks, including biased or incomplete data producing unfair outcomes, reduced human judgment for decisions with serious, life-affecting consequences, and accountability gaps when something goes wrong — serious enough risks that responsible humanitarian organizations generally maintain meaningful human oversight rather than fully automating these consequential decisions.

Go deeper

• What Business Decisions Should Never Be Fully Automated With AI?
• What Are Legal Aid Organizations Doing With AI Chatbots?