AI based risk predictions for child protective services may reinforce Indigenous disadvantage

Embargoed until: Publicly released: 2021-03-29 05:00

Australia; VIC

What does this mean?

Jump to: MultimediaAttachmentsExpert Reaction

Machine learning algorithms, trained on historical data are being used in child welfare agencies to assess the risk of adverse outcomes, but Australian experts warn that these systems may just reinforce the status quo for Indigenous Australians already subject to overrepresentation and over-surveillance in welfare settings. The study found there is a lack of empirical evidence addressing issues relating to racial bias and this sort of predictive risk modelling in child protection, particularly regarding Indigenous peoples.