1

A study reported in Nature Human Behaviour in 2016 generated a lot of media hype. Typical headlines asserted:

The disturbingly accurate brain science that identifies potential criminals while they’re still toddlers

or:

Future criminals revealed at three, says study

The article under the first headline explains the basis like this:

In the study ... researchers led by neuroscientists at Duke University showed that those with the lowest 20% brain health results aged three went on to commit more than 80% of crimes as adults. The research used data from a New Zealand longitudinal study of more than 1,000 people from birth in the early 1970s until they reached 38 years old.

This is more nuanced than the headline but to what extent is the headline justified?

Can we really predict future problems in people by studying their cognitive function as toddlers?

  • You have examined both the article and the reporting on it, are not looking for any more evidence and further debate requires expert subject matter opinion. I suggest we move this to Psychology & Neuroscience. – Sklivvz Dec 31 '16 at 10:45
  • @Sklivvz I asked it here because the newspaper headlines were widespread. I'm not interested in the cognitive science technicalities: i'm interested in leaving an answer here that addresses the headlines (which is one of the key functions of this site). I'm worried that your grounds for migration ultimately apply to every question on this site. And, there is already a useful answer addressing the headlines. – matt_black Dec 31 '16 at 12:24
3

It depends on what the meaning of the word "is" "predict" is.

Let's assume that the study is true (I haven't seen any contradicting ones).

  • Number of prisoners in Australia:

    The national imprisonment rate was 208 prisoners per 100,000 adult population, an increase of 6% from 196 prisoners per 100,000 adult population in 2015 (src).

    Clearly, this isn't 100% same as number of criminals, but still, good enough for government work an estimate below - and the study used "criminal convictions" as a proxy for # of criminals anyway, so we aren't changing methodology here

  • 20% cohort that the study mentions would be 20,000 people out of 100,000.

Therefore, all this study shows is that roughly, of those 20,000 people, 162 (80% of 208) are likely to be convicted of crimes; and 19,838, or 99.2%, are not likely to be convicted of crimes.

In other words, no, you can't predict future problems with people to any degree of accuracy by studying their belonging to lower 20% cohort of brain function as this study did.

Please note that the exact rate of criminals is not important - the picture is only slightly more "predictive" even if you take the much much higher incarceration rate in USA (698 incarcerated rate) - you still get 97.21% of the 20% cohort are not likely to be incarcerated.


What you probably can meaningfully do is to concentrate your preventative measures (efforts aimed at helping what's called "at risk" youth) on the 20% cohort, because there ROI would be slightly higher.

  • "Criminal" is also poorly defined. Many people - I'm tempted to say virtually everyone - commits "crimes" (defined as things that are illegal) of one sort or another, but comparatively few get arrested, and still fewer are convicted. And on the flip side, many people are convicted of crimes they didn't commit. So about all this prediction could manage is to identify those unlikely to avoid being convicted. – jamesqf Dec 30 '16 at 19:35

Not the answer you're looking for? Browse other questions tagged .