I am curious as to what our Sig Forum LEOs think about this program. It is a fairly indepth story but worth watching. I certainly do not see this as violating civil rights, although some do. To me it seems like good police work.

http://wgntv.com/2017/05/25/cp...ategic-subject-list/

Some years ago, I attended a training seminar entitled "Precursor Crimes"
The gist of this was:
If a person (especially as a juvenile) committed certain low level crimes, he / she were then more likely to move up serious felony crime.
Examples:
Voyeurism would indicate a sex offender.
Animal abuse could signal crimes of violence.
CPD is evidently doing the type of targeting offenders, but stepping it up with the use of computers.
After I left the training course, I felt it had merit. But when I attended, there was no real means for long term offender tracking. So I felt the concept was of limited use.
I don't see a civil rights problem, as long as it is used to track known repeat offenders.

This message has been edited. Last edited by: YooperSigs,

Another boondoggle by Jonathan Lewin. He's been a house mouse for years. IMO he needs to spend some time on the street. I predict it will turnout to be another failure and will cost the city some big bucks.

That article doesn't provide much information. Are they just making a list of people who they think are more likely to commit a crime based on their algorithm and then using heightened surveillance on those people to try to catch them in the act?

If so I have to believe it's a pretty big list in Chicago and they don't have the resources to fully devote to the project. I could be wrong though ...

Every cop knows who the punks are on their beat who are on the fast track to Hell, don't need a computer algorithm for that. For years we have had programs that focused police resources on habitual offenders, like VICAP, SHODI, SHOCAP, etc, that are designed to take repeat offenders off the street as quickly as possible.

This algorithm theory is interesting from an academic point of view, but the moment government begins taking adverse action against individuals based upon the theory, it will certainly be squelched by the courts on Constitutional grounds. It may be good information to know but how far do you go with it? The real danger here is that computers become the substitute for critical thinking in policing.

Similar concepts of trait criminology are not new. Phrenology was once advanced as a science that could predict a propensity for crime by the shape and nature of bumps on one's head. Somatotype was another theory that suggested a person's body type predicted criminal behavior. It was once thought that witches float in water and innocents sink. There have been many other theories as well that have been debunked over the years.

The truth is, human behavior is so complex that any attempt to pigeonhole individuals will be faulty. The son of a crack whore with no positive male role models who lives in the hood where violence is common and drugs are currency is at risk of a life of crime, but he might go on to become a prominent surgeon or military officer. Our Constitution will not tolerate adverse action being taken against individuals based on predictive models.

Been around for years.

I'm not an LEO, but I am in a data sciences field. Analytics can predict damn near anything (what movie you want to watch, if you're going to develop type II diabetes) tied to patterns (human behavior) assuming it has enough data elements (age, income, zip code, family diagnoses, or... did Mark123 watch "other" Jason Bateman films).

With enough info, it'd be very reliable. Not reliable enough to arrest, but I think easily reliable enough to be of interest.

quote:

Originally posted by HayesGreener:human behavior is so complex that any attempt to pigeonhole individuals will be faulty.

Agreed, because of the word individual. But, an attempt to identify a suspect sub-population from the population on a whole could be very powerful. Analytics is in it's infancy and is already spectacularly good.

I have a little nightmare that is in my head that one day, the Internet will be a tool to rule us, believing we as a population have free will. I suppose it happens already. With machine learning, analytics and individual message targeting, it will be incredibly effective. I digress...

quote:

Originally posted by Palm:

Are they just making a list of people who they think are more likely to commit a crime based on their algorithm and then using heightened surveillance on those people to try to catch them in the act?

The Machine from Person of Interest?

Rubbish!

As Yogi Berra put it, "my forecasting is none too good, particularly when it pertains to the future."

67.3% of statistics are misleading. The rest are flat wrong.

These "analytics" are useful when analysing mechanical processes, fair coins, unrigged decks of cards. They are far less so when human behavior is involved.

The most spectacular example so far has been Long Term Capital Management, a huge hedge fund founded by some very experienced, very successful traders from Salomon Brothers in the mid-90's. They were joined by a couple of Nobel Prize recipients in economics. Using sophisticatd real time anslytics, they were trading all sorts of instruments, picking up small price disparities.

For several years the scheme worked yielding impressive gains. The chance of a drawdown of even 20% was minescule, the chance of a wipe out something like once in the life of the universe times 7. Then, in a few months, interest rates started gyrating, the Russian bond went sideways, etc, and they burned through something like $5 billion in a few months. The Fed organized a "rescue" to avoid a number of big banks from being wiped out in the collapse.

Analytics might show that 90% chance of folks in a certain group being criminals. It can't tell you which ones.

quote:

Originally posted by JALLEN:

67.3% of statistics are misleading. The rest are flat wrong.

That's incorrect.

The other 43.7% are wrong.

It's just profiling....which everyone but scatacephalics knows works. 'Algorithms ' is how to describe it without empty heads exploding.

Predictive policing. Ah, good luck with that. How about predictive judging, that's where the problem lies.

This applies to adults as well. I remember reading company reports at General Research on Criminal Justice Projects that people committing minor crimes, later moved to major crimes.

quote:

Originally posted by JALLEN:
Rubbish!

As Yogi Berra put it, "my forecasting is none too good, particularly when it pertains to the future."

67.3% of statistics are misleading. The rest are flat wrong.

These "analytics" are useful when analysing mechanical processes, fair coins, unrigged decks of cards. They are far less so when human behavior is involved.

The most spectacular example so far has been Long Term Capital Management, a huge hedge fund founded by some very experienced, very successful traders from Salomon Brothers in the mid-90's. They were joined by a couple of Nobel Prize recipients in economics. Using sophisticatd real time anslytics, they were trading all sorts of instruments, picking up small price disparities.

For several years the scheme worked yielding impressive gains. The chance of a drawdown of even 20% was minescule, the chance of a wipe out something like once in the life of the universe times 7. Then, in a few months, interest rates started gyrating, the Russian bond went sideways, etc, and they burned through something like $5 billion in a few months. The Fed organized a "rescue" to avoid a number of big banks from being wiped out in the collapse.

Analytics might show that 90% chance of folks in a certain group being criminals. It can't tell you which ones.

The mid 90s was a long time ago. FaceBook, Google and Amazon and a thousand other companies are based upon predicting a myriad of outcomes from behavior and of behaviour. It's not rubbish, whether you believe it or not.

The average beat cop that has high standards and a good work ethic doesn't give a rats ass about predictability studies. Bad guys generally begat bad guys in the hood. Shocker. Patrol is the backbone of ANY city department with any crime problem to speak of. Getting trustworthy qualified individuals to apply to be police officers is a much bigger issue to PD's than crime predictability studies.

I just wear Sex Panther. 50% of the time, it works every time.

Here's an idea...when the cops lock someone up for a crime, prosecute them and keep them in jail.

The shitheads that do really bad stuff have almost ALWAYS already been arrested for major offenses, and were let off. Usually completely let off. And if the DA manages to prosecute, they accept a plea for a few months or only probation.

It is, truly, all just a joke. Don't try to pretend otherwise. Rarely, one will get real time the first time, for killing an innocent victim that made real headlines.

Otherwise, no matter what the police do, the DA, the judge, etc., are going to shrug it off. I could tell you mind boggling stories. Expect nothing, and be surprised when you get something.

This is precisely why bad people are locked up for decades on drug charges. The courts won't do shit to them for shooting or robbing a few people, or can't because of a lack of witnesses, but hang 20 drug buys on Johnny's neck with video, accumulated weight, and a gun...nobody can sweep that under the rug, and Johnny goes to club fed for 20 years.

Its my understanding that another facet of predictive policing involves predicting where the next crimes will occur. Like when they see a rash of car breakins, they can pretty well assume there will be more in teh same region. Of course seasoned detectives will just tell you the same thing, but having it come out of a computer is more believable. It can even suggest the same unknown criminals will be involved. So you can deploy extra manpower to watch.

And sometimes the courts get involved when the cops get a hunch and want to stake out one bad guy or group. Not fair to watch the guys who break in to houses when they go out at night, or during the day.

Maybe we can have more episodes of Bait Car with predictive policing. I could think of other entertaining programs that skirt the edge of entrapment.