Software To Predict Who's Likely To Be A Murderer

from the minority-report,-here-we-come dept

Since the movie Minority Report (based on a Philip K. Dick short story) came out in 2002, we’ve seen a number of technology advances in that movie actually show up, in such things as posters that talk to you. However, much more interesting was the central tenet of the story, about police being able to predict crimes. We’ve seen some basic reports that try to predict how much crime is likely to happen in a certain area, to figure out how to better deploy police. Then, last year, there was the wholly unbelievable story claiming that a computer system predicted a mugging, though details were lacking. Now, we’ve got Bruce Schneier pointing to a story saying that a researcher at UPenn is developing software for the city of Philadelphia that will try to forecast who is likely to become a murderer out of criminals already in the probation department’s computers. Obviously, the idea isn’t to lock these people up for a crime they haven’t committed (er… we hope), but to put extra emphasis on working with those individuals to try to keep them from fulfilling the prediction. Certainly an intriguing (and some may find, worrisome) idea — but it would seem like there would need to be an awful lot of proof to find out if such software can be even the tiniest bit accurate. Also, though, you have to wonder if part of the reason why this story catches your attention is because of the use of the very specific crime of murder in the description of the software. Would the software be as interesting if it was described as just predicting which current criminals were at “high risk” of later committing a more serious crime?

Rate this comment as insightful
Rate this comment as funny
You have rated this comment as insightful
You have rated this comment as funny
Flag this comment as abusive/trolling/spam
You have flagged this comment
The first word has already been claimed
The last word has already been claimed
Insightful Lightbulb icon Funny Laughing icon Abusive/trolling/spam Flag icon Insightful badge Lightbulb icon Funny badge Laughing icon Comments icon

Comments on “Software To Predict Who's Likely To Be A Murderer”

Subscribe: RSS Leave a comment
Anon says:

Re: the king

Oedipus consequences? They are going to kill their father and sleep with their mother???

“what are the chances that these additional stresses could lead them to cause a murder”
Sounds more like self fulfilling prophecy…

Also, I agree with Blah, there is no way to prove that it works. If the program identifies a bunch of people as “likely or potential” murderers and those people then do NOT commit murder, they would be able to say “See, the program worked and because of intervention we prevented murder”????

Blah says:

Fundamental flaw

If this thing works they’ll have a really big problem on their hands.

A) If they successfully predict the right people, and are able to help them and prevent them from becoming murderers, it would be impossible to tell if they were in fact going to be murderers.
B) If they find people they think will be murderers, give them counseling, and those people still become murderers, you can’t prove they’d have been murderers without the counseling because of (essentially) the Heisenberg Uncertainty Principle.
C) In any other case, they’ll either be preventing murders that wouldn’t have happened anyway or causing murders that otherwise wouldn’t have happened. In either case, there would be know way of knowing what would have happened absent this system.

The only way to really test the system is to make a very private, well-protected list of these individuals and see if they ever do in fact commit murder; the problem here is that as soon as politicians who are out of power hear that murders were predicted and not prevented they’ll attack the whole program. My prediction: They’ll use the system, and base their success on whether murder rates go down, regardless of whether any change in murder rate is in fact related to this system.

chaalz says:

Re: Fundamental flaw

Just playing along for a sec, if the program produced a list of 1000 people, you could “treat” 500 of them and not the rest and then compare the rates. Obviously 1000 is a small number but over time as that number grew, one could make an argument for or against the system based on the results.

What would you say if only 25 of the 500 that were “treated” went on to commit the crime, but 350 of the 500 that were not “treated” went on to commit the crime. What if you did for for 10,000 or 100,000 people? hmm…interesting stuff.

Anonymous Coward says:

Re: Fundamental flaw

My prediction: They’ll use the system, and base their success on whether murder rates go down, regardless of whether any change in murder rate is in fact related to this system.

My prediction: They’ll use the system, and if the murder rate goes down they’ll credit the system. If the rate goes up, they’ll say it would have gone up more if not for the system. In either case, they’ll claim success and that the system needs to expanded to other crimes.

Ajax 4Hire (profile) says:

The police and public do it now...

its the note that a software program is falling out the names that gets people worried. Do not be distracted by the fact that hardware/software point the finger.

It is people who write the software, compile the database, run the queries and act on the results.

And besides, police have been doing this sort of thing as long as police have been around: “round up the usual suspects.”

There is nothing special about this except that someone has had the forsight to load all of the personality, societal, economic data into a searchable database and found correlation.

The staticitian will tell you that correlation does not prove causality;

As humans you look for patterns and act on those patterns. I don’t go pet the lion because I am prejudiced that he will eat me; I have prejudged the lions intent. Until lions show a little more compassion and true heart felt change, I will be wary.

redhammy says:

they probably already have test data

I bet they are probably using a typical Data Mining/Machine Learning approach, where they have a large set of historical data with statistics about people who did and did not eventually commit murders. They will use a portion (say, half) of the data to “train” the system. Then, they will use the rest of the data to test the system and get accuracy statistics like precision and recall. I don’t really know if that’s what they are doing but this is a pretty standard approach.

Anonymous Coward says:

software only does what it is told. sure, the software can “learn” but what it is actually doing is readjusting its logic/math to better fit the results. i had a class on neural networking, (i.e. making computer simulations of how the brain works) and we tried to predict things, and make computers learned. however, due to large ammounts of data, variables/results. it is very difficult to get a percise measurement. and this was with objects that don’t have “natural control” over them.

what i’m getting at, is that we don’t know the human brain. we don’t know the transfer function for the human brain. we don’t know what pre-triggers will lead to what end-results. just because 2 people have the same pre-trigger, doesn’t mean they’ll get the same end-result.

sure this is a good expierement, however i don’t think it should be used to “protect” the people. remember, the police are there to ENFORCE the laws, not prevent them. sure they take a role in trying to prevent criminals as much as possible, but in the end, they are there AFTER the crime has been committed.

Anon2 says:

George Orwell write this?

Technically, yes he did. He warned us of the way we were heading. The New World Order is upon us and making the final moves to take control. Look at what the European Union is doing to Europe. [Google Video – 50 min]

Bush and the leaders of Canada and Mexico recently met in Canada and signed an agreement that will form the Nort American Union. Don’t think it’s just a Neo-Con thing either, partisan politics is only a smokescreen over all the corruption throughout our Federal Government. America: Freedom to Fascism[Google Video – 1 hr 50 min]

The proof is all around you and making headlines every day. Before you cry “conspiracy theory” remember that many actual conspiracies preceded recent “theories.” And don’t forget the National ID that was predicted in the Bible.

Anonymous Coward says:

I agree with the above, we don’t know how the brain works. We can’t assume we know how others are going to act, and program a computer based on those assumptions. There’s no true way to eliminate biases either.

You could program it based on statistics and a “basic case” study of crime- but how long will that take? You’d have to do it for each region, each county- in each city, in each state. Because I can promise you, the crime rates in Buffalo, New York, would be a lot different compared to the crime rates in Miama, Floridia.

There would be too many flaws in the system (not that our justice system isn’t already flawed), and there’s way too much of an ethical debate. You talk of the murder victims’ families, but what of the accused murderers’ families?

And that’s not even counting the fact that we’re low on police officers and detectives- creating a new unit, something that would specailize only in pre-cog crimes, would cause more of a deprevation of our officers on the streets.

Tyshaun says:

A few issues...

I would have a few issues with this:

1. Basically, utilizing this system on parolees is like saying that there is no faith in the criminal justice system. The point of incarceration is both punishment and to change criminal behavior patterns. By using this system on parolees we are saying “yeah, we know the jails don’t work at changing criminal behavior or keeping those that would do us harm out of society”.

2. I’m not a lawyer but this seems to be dangerously close to a judgement without a trial issue, especially if those people spit out on a list are somehow forced to undergo some type of treatment based on the computer assessment. What would be the penalty for non-cooperation? Is there some type of way to challenge the decision?

3. Wouldn’t the government now be liable in some way if they have a list of those likely to commit a murder and those people still commit murders? There was a case near me, wish I could find a link, where a child molester was released from jail and his parole officer didn’t verify he registered with the local police as a child molester (which is the law). The person went out and molested a little girl who lived next to him. The parents sued the city and state and won. I can just imagine a similar situation.

4. As was noted, what about the human factor? Let’s say there is some type of increased monitoring or worse for “potential offenders”. What is to prevent malicious attempts to add people to the list by altering data or even program code?

redhammy says:

hang on

If they have some treatment program that reduces a convict’s likelihood of committing a murder, then they should give that treatment to as many convicts as possible, should they not?

Since resources are limited, they are going to have to give out the treatment to those who, statistically, are more likely to commit murder. Given enough training data, this statistical analysis is exactly what computers and computer scientists are good at, despite what difficulties anybody may have had in their neural net class project.

Nobody is saying it will be 100% accurate, and nobody is saying they are going to find anybody guilty of anything based on this.

Anonymous Coward says:

another thing. there is misconception that once you “leave jail” you are free. well, your not free. felons can’t vote or own guns. they may have to register as an ex-con (or sexual assult) they may have to put up with harrasment from police due to being an “ex-con”. so once people leave jail, they still haven’t paid their debt to society.

why not, instead, just keep everyone in jail? i mean, the only people these criminals would murder would be other criminals. and hey, the less criminals, the better. do what england did, ship the criminals off to some remote island, and let them live there. (cruel and unusual? maybenot. i mean, they are keeping the streets safer.)

Add Your Comment

Your email address will not be published. Required fields are marked *

Have a Techdirt Account? Sign in now. Want one? Register here

Comment Options:

Make this the or (get credits or sign in to see balance) what's this?

What's this?

Techdirt community members with Techdirt Credits can spotlight a comment as either the "First Word" or "Last Word" on a particular comment thread. Credits can be purchased at the Techdirt Insider Shop »

Follow Techdirt

Techdirt Daily Newsletter

Techdirt Deals
Techdirt Insider Discord
The latest chatter on the Techdirt Insider Discord channel...