First, Do No Harm
MetroLab Network released a new report providing brief practical guidance to local governments on mitigating the risks of applying data science techniques to human services. The report is structured around four principles: engagement, model validation, review, and transparency. Each principle is paired with specific cautions and recommendations as well as suggestions about how to approach foundational issues like data stewardship and crisis communications. While numbers are objective, NCCD CEO Kathy Park also notes, “To do this work, you have to acknowledge the biased nature of administrative data. Administrative data reflect both structural and institutional bias and are also influenced by individual cognitive bias, which drives certain families into the system more than others.” Whether predictive models reduce or amplify the structural biases in the sector may depend on how vigilantly local governments and child advocates adopt and improve upon ethical guidelines like these.