Jamie Fuller: Much Distress and Little Relief

In last week’s Open Studio presentation, Jamie Fuller discussed the dangers of using AI-driven “predictive optimisation” in South Africa’s Social Relief of Distress (SRD) grant system. Fuller, a  junior researcher at Research ICT Africa, explained how the new automated system was introduced to combat fraud by requiring grant recipients to verify their identity through a selfie upload. However, the technology inadvertently excluded thousands of vulnerable citizens from receiving essential support due to issues like lack of access to smartphones, poor internet, and outdated ID documents. For many of South Africa’s poorest, the process was simply inaccessible, leaving them without the financial help they rely on.

Fuller critiqued the decision to prioritize fraud detection over ensuring broader access to the grant, especially given South Africa’s digital inequalities. While AI may seem like a quick fix, it often fails to account for the socio-economic challenges faced by marginalized populations. In this case, the reliance on digital systems compounded existing inequalities and excluded those unable to meet the technological demands.

At the end of the Open Studio, Fuller emphasized that while AI has the potential to improve welfare systems, it must be deployed carefully, with transparency and a clear understanding of the context. She highlighted some potential areas where predictive optimization could be used to address inequalities, such as identifying those who are missing out on benefits, not just detecting fraud. However, given the existing transparency and accountability issues with the predictive optimization system, she warned not using it for other purposes until the underlying problems can be more fully addressed.