Helping professionals are often called upon to make predictions. Child protection workers are asked to predict which parents will continue to abuse their children, correctional authorities are asked to predict which offenders will re-offend and domestic violence worker – which perpetrators will continue to assault their partners and, some cases, will murder those partners.
But inevitably these predictions are wrong.
Let me illustrate with reference to prediction of suicides. https://www.youtube.com/watch?v=6QPI0AFQ_Yw
In 2014, there were 42,773 deaths by suicide in United States. Further there were some 1.1 million serious attempts by people to take their lives. That is there were 26 serious attempted suicides for every successful suicide.
Further, epidemiological surveys show that 9.1 million people had serious thoughts of suicide during 2014. That is for every one person who suicides. 220 people thought seriously about taking their lives.
42,773 suicides represent 0.0001% of the US population in 2024. A very low base rate.
Thoughts of suicide and attempts at suicide are referred to as ‘risk factors ‘for suicide ie it is likely that people who suicide have made unsuccessful attempts in the past and have constant thought of taking their own life. But there are by themselves poor predictors thus my argument that the helping professions ‘suck ‘at predictions.
An expanded version of this blog appears on my blog where I discuss the implications for the media.
The Media
The media are fond of pointing out that when tragic events occur such as a serious crime or a child is killed by their parents that the perpetrator is ‘known to police or ‘is known to child protection authorities’. The implication is that these events could have been prevented. But could they have been?
A simplified example
Let’s assume that a risk factor for a child being killed by a parent is that the parent has a serious substance abuse problem. But the reverse is not true ie all parents who have addiction problem kill their child.
If we remove all children from parents who have serious addition issues, then it is likely that no children will be killed by their parents. But the cost is many (thousands?) of children would be removed unnecessarily from the care of their parents.
Here is the rub. If we use risk factors to predict tragic outcomes almost inevitably, we will sometimes get it wrong. Some people will take their life just as some children will continue to be abused, some offenders will continue to offend and some DV perpetrators will continue to assault their partners. There is a solution but involves punishing/incarcerating or otherwise curtailing the liberties of people based on a prediction of what they might do in the future.
Surely this is a recipe for increasing the number of people in our jails.
I am not suggesting that we accept the status quo. We should conduct a rigorous analysis of the lead up to any tragic event but be aware that it is possible that all people involved have followed required processes and no-one is to blame.
