Probabilistic Crime Solving

You may also like...

1 Response

  1. Ipso Facto says:

    Santayana– “Those who cannot remember the past are condemned to repeat it”

    “Risk terrain modeling” reminds me of other buzzwords like “defensible spaces.”

    I’m reminded because I’ve read it all before (and I’m only in my 40’s). Analyses of crime versus terrain (built or otherwise), and crime versus all sorts of factors, have been done to death since the 1960’s, as a few minutes in the library will confirm.

    The grant to Rutgers is pure pork because the Rutgers work has only a miniscule probability of discovering anything novel or interesting. I would guess that the situation is worse than wasteful– the grant is probably part of a campaign to conceal, rather than reveal, the essentials about crime, which is that people, not places, commit crime, so the strongest predictors of criminal activity are personal characteristics.

    “Cluster analysis” shows that people can be grouped by certain variables, that the resulting groups have noticeably different mean and std-dev propensity to commit various sorts of crime (e.g.g., drunkenness, violence, larceny, fraud), and that some interventions are useful (e.g., limiting access to alcohol) and some are not (e.g., moving adults from “bad” neighborhoods to “good” ones– when you do that, you simply relocate the crime along with the people, because people–not neighborhoods–commit crime).

    If “predictive policing” means “concentrating police effort in places where people with a high propensity to commit crime are found,” then I’m for it. If it means “spying on everyone all the time” then I’m against it.