What's Happening?
An investigation by The New Humanitarian has revealed significant flaws in GiveDirectly's AI-powered flood relief program in Ibaji, Nigeria, launched in 2024. While GiveDirectly initially reported that 4,600 individuals received anticipatory aid, leading
to increased incomes and reduced food insecurity, the investigation found that many of the area's most vulnerable residents were excluded. Reasons for exclusion included personal data not matching government or bank records, and the inability to afford phone charging to receive verification calls. Some excluded farmers incurred debt expecting aid that never materialized. Conversely, the program's verification and targeting systems allowed some aid payments to go to relatively wealthy individuals and those living outside the designated flood-affected areas, sometimes with multiple members of the same family receiving separate payouts. GiveDirectly acknowledged that the program's design inherently excluded some vulnerable groups, such as those with limited phone access or who had moved to higher ground before the program began.
Why It's Important?
This situation highlights critical challenges in the implementation of technology-driven humanitarian aid, particularly the potential for AI and digital systems to exacerbate existing inequalities and exclude the most vulnerable populations. The reliance on digital verification and communication methods in areas with limited infrastructure, such as unreliable electricity and mobile connectivity, creates significant barriers for those who need aid most. The exclusion of individuals due to data mismatches underscores the digital divide and the complexities of identity verification in developing regions. Furthermore, the misdirection of aid to wealthier individuals or those outside target areas raises questions about the effectiveness and ethical implications of AI algorithms in humanitarian contexts, potentially undermining trust in aid organizations and their methods. This case serves as a cautionary tale for the broader humanitarian sector as it increasingly adopts technological solutions.
What's Next?
GiveDirectly has stated it is incorporating lessons learned from the Nigeria pilot into future work and is studying similar anticipatory action models in Kenya, Mozambique, and Bangladesh, including a large randomized controlled trial covering over 100,000 families in the Jamuna River Basin. The organization is actively working to improve its processes for addressing exclusions and misunderstandings, including conducting 're-sensitisation visits ward by ward' to clarify program details. However, the long-term impact on trust within affected communities remains a concern, especially for those who were excluded without clear explanation and incurred debt. Future implementations of AI-powered aid will need to prioritize robust, inclusive verification methods and address infrastructure limitations to ensure equitable distribution and prevent further marginalization of vulnerable populations.
Beyond the Headlines
The incident in Ibaji exposes a deeper ethical dilemma in the pursuit of efficiency through technology in humanitarian aid. While AI promises faster and broader reach, it risks dehumanizing the aid process by reducing complex human needs to data points. The 'deliberate tradeoff' of speed and coverage over precision targeting, as described by GiveDirectly, raises questions about accountability and the true cost of such compromises. The reliance on digital identity and communication systems can inadvertently create new forms of exclusion, particularly for those already marginalized by poverty, lack of education, or geographic isolation. This case underscores the need for a human-centered approach to humanitarian technology, where algorithms are designed with a profound understanding of local contexts, vulnerabilities, and the potential for unintended consequences, ensuring that technological advancements genuinely serve, rather than bypass, the most in need.
