Better Together? A Field Experiment on Human-Algorithm Interaction in Child Protection

Marie-Pascale Grimon and Christopher Mills

Papers from arXiv.org

Abstract: Despite algorithms' potential to improve public service efficiency, lack of evidence and widespread concerns about equity have impeded their deployment. We conduct a large-scale randomized controlled trial ($N=4,681$) where we provide algorithm support to workers allocating Child Protective Services (CPS) investigations in real time. Relative to humans on their own, access to the algorithm reduced maltreatment-related hospitalizations, including a 29 percent reduction in child injury admissions. Algorithm support also improved equity, reducing harm disproportionately for historically disadvantaged groups, while reducing CPS surveillance of Black children. We further show that humans with algorithm support outperformed the algorithm on its own, using counterfactual exercises designed to be applicable in settings where an algorithm-only treatment arm is infeasible. Finally, we provide suggestive evidence using novel LLM-based methods applied to worker discussion notes on how workers reallocated investigations to children at greater likelihood of harm, irrespective of algorithm-predicted risk.

Date: 2025-02, Revised 2025-08
New Economics Papers: this item is included in nep-exp
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://arxiv.org/pdf/2502.08501 Latest version (application/pdf)

Related works:
Working Paper: Better Together? A Field Experiment on Human-Algorithm Interaction in Child Protection (2025)
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:arx:papers:2502.08501

Access Statistics for this paper

More papers in Papers from arXiv.org
Bibliographic data for series maintained by arXiv administrators ().