Predicting less, understanding more: shifting the use of machine learning from individual prediction to structural insights in systems that affect children and families

Machine learning algorithms (MLAs) have been deeply embedded across domains, valued for their capacity to analyze large-scale data and their ability to support a range of descriptive and predictive tasks. Despite their versatility, most applications of MLAs to date have focused on individual-level prediction at the expense of broader structural insights. This paper shifts that focus by using MLAs to examine how juvenile and family court intake structures shape case outcomes in delinquency referrals. Specifically, we apply machine learning to evaluate how the structure of juvenile court intake decision-making affects juvenile referral outcomes. In doing so, we offer a novel demonstration of how these tools can support causal inquiry into the functioning of complex legal systems. Results indicate that juvenile court intake structure has a material effect on youth outcomes in the system. That is, when prosecutors act without input from the juvenile justice agency, youth referrals are significantly more likely to be prosecuted rather than diverted or dismissed. Differences in prosecution rates between prosecutor-controlled intake (PCI) and agency-involved intake (AII) structures are most pronounced among Black girls and least pronounced among White boys, highlighting how system design choices may contribute to racialized and gendered disparities in juvenile justice outcomes. Taken together, these findings underscore the potential of MLAs not just to predict individual outcomes, but to generate insight into the policies and practices that shape them—opening new directions for structural analysis and reform in systems that affect children and families.