ABSTRACT
In 2023, Rotterdam discontinued an invasive, biased welfare fraud risk-scoring algorithm after an investigative report by Lighthouse Reports, which exposed its racial and gender biases, disproportionately affecting migrant mothers in deprived areas. This chapter argues that such biases could have been identified before implementation by scrutinizing the categories embedded in the algorithm and contextualizing them within the history of the Dutch welfare system. Using a genealogical approach, we trace how norms about race and gender became embedded in welfare practices. A category analysis shows how these biases shaped the algorithm’s indicators. Drawing on critical data studies and feminist theories on migrant motherhood and racialized citizenship, we show how discriminatory ideas about the “ideal” welfare recipient predate the algorithm, contributing to discussions about equality in datafied welfare governance.
