Right now, the police can't do much to help you until after a crime has been committed. In a science-fiction tale about free will and psychological determinism, Philip K. Dick's "Minority Report" detailed a world where PreCrime officers stop acts before they happen. They know who is going to do it.

But psychic mutants are a simplistic convention, criminologists have instead been trying to do it in the real world.

No matter what you read in cultural frothing-at-the-mouth about psychotropic medicines, economic inequality or gun control, murder is actually a rare crime, it happens to 1 in every 100,000 people overal - but there are high risk groups where it is 1 in 100. So a few years ago, Richard A. Berk. Professor of Criminology and Statistics at University of Pennsylvania, and colleagues got together and created an algorithm that sought to predict who was going to kill or be killed.

It was narrow, using people on parole or probation, but they found their their algorithm could predict a subset of people much more likely to commit homicide when on parole or probation. Instead of finding 1 future murderer in 100, they could identify 8. It has been tested in crime-ridden cities like Baltimore and Philadelphia.

The two biggest factors, they found, were the type of crime and the age at which it was committed.

Dr. Alex Piquero of The University of Texas at Dallas and colleagues instead say there are 5 factors that distinguish people who commit homicide from those who don't. They used data from Pathways to Desistance and found that 8 demographic characteristics and 35 risk factors distinguished the 18 juveniles charged with homicide from those who were not from among the 1,354 youths who were charged with serious crimes.

Like with Berk earlier, they found that one demographic measure - age - was important. There were 4 other commonalities in people who committed homicides, some of which are too subjective to be prospective rather than retrospective. Prospective is what we need. Anyone can accurately predict the past. They are:

  •  Significantly lower IQ

  •  Violent neighborhood 

  • Higher perceptions of living in a neighborhood characterized by disorder

  •  Higher prevalence of gun-carrying

But of the 5 factors only 2 were statistically significant, and it wasn't something obvious, like owning a gun or age; Piquero and colleagues say instead the biggest risks were IQ and living in violent neighborhoods.

Everyone knows that violence occurs more in some neighborhoods than others - whether or not criminals turn normal communities into violent ones or violent neighborhoods turn people into criminals is a sociological debate but not a practical one; anyone who has had a bad neighbor move next door knows the answer. What that 'tipping point' is, where criminal behavior becomes acceptable in a neighborhood, is unclear - studies have also found that in some wealthy urban settings and on crowded city streets, people ignore crime - we have been trained not to get involved and a new generation has been trained to capture it using photojournalism but not do anything. And some sociologists have given criminals a free pass in recent decades, making violence egalitarian and exculpatory by blaming income or schools, but the new paper isn't trying to do that, it just wants to dispel the belief that murderers are psychopaths or mentally ill, which, along with gun availability, has been the default explanation for things like school shootings in recent years.

Homicide isn't deterministic, based on age or income, they say, but rather an anti-social behavior. Prevent anti-social behavior and you prevent crime, a different approach than a city like Baltimore does, where they monitor criminals who fit the profile more closely.

And it avoids implications of harassment of people who are convicted felons but haven't done anything new.

Large police departments aren't going to accept something as sociologically fuzzy as trying to prevent neighborhoods from being "anti-social", they have to live in the real world - and that means predictive policing. The Real-Time Analysis and Critical Response Division of the Los Angeles Police Department watches for a sudden surge in criminal acts in any neighborhood, because that means it could be about to become a tipping point. Once criminals find a place that is an easy target, they go back. The LAPD uses software called PredPol, developed with input from anthropologists, psychologists, crime experts and statistics experts.<

Unlike Berk's algorithm, which focuses on people, or Piquero's, which focuses on sociology, PredPol predicts violent hot spots, something even the most experienced police officer would have trouble doing over time. The LAPD's single-blind randomized control trial found it was 2X as accurate as legacy statistical models. The United Kingdom countries of England, Scotland and Wales lead the developed world in violent crime and as a result some cities there have also adopted PredPol, so that they can focus in potential hotspots rather than trying sparser blanket coverage. Criminologists don't believe that when more police enter an area, the criminals move somewhere else, they believe it lowers crime.

The risk is always going to be that someone could be unfairly targeted, but that happens now. The first thing police do when a crime occurs is look at The Usual Suspects, just like if a wife is murdered they focus on the husband; statistics show that in an alarming number of cases, that is true.

The times it isn't true are rare enough that television shows and movies like "The Fugitive" get made about them.

 Predictive Policing Image: PredPol