Predicting Crimes

February 28, 2012 in Daily Bulletin

In The Atlantic Nadya Labi looks at a new type of statistical analysis that is taking the criminal justice system by storm: algorithms that predict the likelihood of violent crimes. Some of the things she found include:

  • The statistics show that teens are at the highest risk of committing another crime after they have been convicted for the first time. Violent activity decreases with age through the 20s, but reoffenders become more common among those in their late 30s.
  • The nature of the crime for which an individual is convicted is not a good predictor of whether or not they will commit a future violent crime. Instead the age and gender of the individual as well as the timing of violent activities matters.
  • While the algorithms do not explicitly take race into account, race still manages to find its way into the equations. Certain statistical models take into account the zip codes of the offenders and these can be divided along racial lines.

To read more about how different cities are using these statistics, what the Supreme Court has to say, as well as some of the ethical issues involved, click here.

Source: The Atlantic

Via: Marginal Revolution