An Ethical and Technical Evaluation of the Use of Machine Learning Models in Health and Human Services: A Case Study of the Allegheny Family Screening Tool

Abstract

In this paper, I conduct a case study on the Allegheny Family Screening Tool (AFST), a risk assessment tool used in child protection services in Allegheny County, Pennsylvania. First, I will review the implementation, use, and impact of the AFST on screening decisions from both a technical and ethical standpoint. Then, I will consider two points of contention in the public debate surrounding the tool: a charge of discrimination against the tool and a confusion about the purpose of the tool. I argue that the tool is not wrongfully discriminatory and that the underlying design of the tool is flawed and must be changed in order to increase accuracy and consistency of screening decisions.