A visual analytics tool helps child welfare specialists understand
machine-learning predictions that can assist them in screening cases.
According to the Centers for Disease Control and Prevention, one out of every seven children in the United States had been abused or neglected in the previous year. Every year, child protective services organizations across the country receive a large number of reports of alleged neglect or abuse (about 4.4 million in 2019). Because there are so many cases, some organizations are using machine learning algorithms to assist child welfare specialists in screening cases and deciding which ones to refer for further investigation.
However, these models are useless if the people they're supposed to help don't comprehend or believe the results. Researchers from MIT and other universities have undertaken a study to discover and address machine learning usability issues in child welfare screening. The researchers evaluated how phone screeners analyse situations with and without the use of machine learning predictions in partnership with a Colorado child protection department. They created a visual analytics tool that employs bar graphs to highlight how certain aspects in a case contribute to the anticipated probability of a kid being removed from their family within two years, based on comments from call screeners.
We'll now discuss how to make machine learning more relevant to high-stakes decision-makers.
www.dprg.co.in
