Author: Elizabeth Brico
Publisher: Talk Poverty
Publication Year: 2019
Summary: The following article discusses an algorithm that is being implemented in Allegheny County, PA that is intended to assist its child welfare programs by predicting which families/parents may be at risk of maltreating a child and intervening with resources. The new algorithm is similar to an old one that was used, and had outcomes that ended up targeting the poor because the data came from social service use. This model uses universally available data (not only data from those with social needs) but the sources being used may have their own issues (juvenile records, homeless services, criminal records, etc.) due to disparities that we have seen in policing and other systems. The concern that “poverty will be confused with neglect” is legitimate, and seems to stem primarily from the data that is being used for this prediction. It may even dissuade individuals from seeking needed help if they think it will have negative externalities later down the line. This article is a comprehensive example of how intentions and outcomes are not aligned, and that collecting adequate data on communities being impacted, with their consent and input, is critical.