July 22, 2024
Cited article by The New York Times
HRRC cautions against the over-reliance on predictive algorithms by law enforcement to determine women’s vulnerability to domestic violence. We recommend governments take protective measures, including conducting independent external audits to detect algorithmic flaws and to ensure human accountability and decision-making in the use of such tools.
News Brief
A recent article from The New York Times (NYT) reports on increasing concerns regarding Spain’s dependence on a predictive algorithm called VioGén to address gender-based violence. While VioGén has contributed to lower repeat abuses committed against domestic violence victims, it has also led to the misclassification of women’s risk levels, where women identified as being in less danger of re-assault were later fatally attacked by their partners. Specifically, NYT discovered that since 2007, at least 55 out of 247 women killed after being evaluated by the software had been assessed as facing negligible or low risk of re-assault, where these deaths are potentially attributable both to classification errors and the fact that VioGén was intended to enable the police to prioritize providing resources including shelter access and monitoring to women at higher risk of being targeted again, while women at lower risk levels receive less protection.
According to NYT, officials state that though instructed to override VioGén’s guidance depending on each incident, Spanish police still abide by the software’s risk assessments in the vast majority of cases, and judges can also utilize risk scores to grant restraining orders. This reliance on algorithmic recommendations can be dangerous especially considering the possible inaccuracy of the information fed into VioGén to produce risk scores. NYT reports that when the police speak to victims to fill out the VioGén questionnaire, fear as well as circumstances such as precarious citizenship status and financial dependence can make victims reluctant to disclose critical information, which has led victim advocacy groups to call for trained experts to direct interviews with victims.