It is a disturbing idea to put into a sentence, but in case after case, America seems to start assembling the warning signs only after a child is dead. The case of Kailey Erhart is a brutal example. The 22-year-old mother, accused of killing two of her children and seriously injuring a third in Mandan, North Dakota, on August 6, had previously been reported by the children’s father as having threatened to kill herself and the children.
Child Protective Services had been notified, and Erhart had undergone psychiatric treatment during the previous 12 to 18 months. Shortly before the alleged attack, cellphone video reportedly showed her carrying a child with a pistol at her hip and telling the father, “Say goodbye to your kids.” Here is the unbearable question that immediately comes to mind as details accumulate and headlines shock: The warnings appeared to be there, so why weren't these kids' deaths prevented? Stats may provide some insight, no matter how disheartening.
In 2024, child-protection agencies across the United States received an estimated 4.365 million referrals alleging abuse or neglect. More than 2 million were screened in for a response, and 532,228 children were ultimately identified as victims. The federal government recorded an estimated 1,773 children who died from maltreatment that year.
Those figures illustrate only a small fragment of the story—the children who died are not simply the fatal fraction of that year’s referrals—but they expose the scale of the prediction problem. Millions of warnings enter systems designed to assess risk every year, and only a tiny share in the kind of catastrophe that makes earlier signs look unmistakable. That is the hindsight trap at the center of the Erhart case.
Institutions must decide what a warning means before they know the outcome, while everyone else gets to judge that decision after the uncertainty has disappeared. The real question is what should happen when several uncertain warnings begin to accumulate at once. The difficulty is that warning signs are not predictions.
A psychiatric hospitalization may raise concern but isn’t a sure sign of what’s to come next. Once a killing occurs, though, the outcome reorganizes everything that’s happened before. Suddenly, the psychiatric hospitalization is seen as a possible prediction.
Extract — continue reading at the source.
