Police Program Aims to Pinpoint Those Most Likely to Commit Crimes
nytimes.com
nytimes.com
http://www.economist.com/blogs/democracyinamerica/2014/03/am...
> This overreliance on imprisonment can be seen most starkly, and sadly, by looking at the juvenile population, which is just under 71,000 nationally. Around 11,600 are imprisoned for "technical violations" of their probation or parole terms, rather than because they committed a new crime. In 11 states such juvenile prisoners outnumber those in for crimes against other people. In only one state (Massachusetts) did juveniles imprisoned for crimes committed against people comprise a majority of juvenile prisoners. Around 3,000 are locked up for things that aren't crimes for adults, "such as running away, truancy and incorrigibility." Incarcerated children are less likely to graduate high-school and more likely to spend time in prison as adults. If America is interested in reducing its prison population, locking up fewer juveniles for silly reasons would be a good place to start.
Yeah and I'm sure that Chicago detectives had absolutely no clue about those 400 people, they really needed a piece of expensive software to tell them where to look for. The trend is to think that technology will solve all problems, but it's just wishful thinking imho.
Wouldn't it be the case that, for example, those in the drug business are far more likely to harm their competition than their market?
And the landlord of the pub where this happened was very lucky a few of the lads didn't go down and make our displeasure known.
> Exactly and as having been in the wrong place at the wrong time this rankles and I did have at the time dissuade my one of my coworkers from "accidentally" putting the perp on the kiddie fiddlers regsister.
I've been a victim of crime - I was in the wrong place, at the wrong time.
It rankles - I wanted revenge.
At the time I dissuaded a colleague from deliberately putting the name perpetrator of the crime on the sex offender register.
Software is expensive to develop, but once developed it is actually very cost-effective. It can be copied over to other departments at a fraction of the cost of a detective salary.
Software will continue to eat the world. Criminals use new technology to stay ahead of the police, so the police has to stay up-to-date too. Data mining software helps the police do their jobs more efficiently and honestly. Factors don't lie, machine learning actively combats bias. While human intuition can be flawed and biased.
There is a danger than humans grant too much authority to computer systems, but there is also an opportunity to remove or dampen cognitive bias.
Given the problems in police departments (which have fortunately started to appearing in the news), giving the police a system that will essentially let them see what you want to see is a terrible idea. Police work is already full of "forensic tools" that don't actually work (like idea that fingerprints actually identify someone uniquely, or the various techniques that are examples of the Birthday Paradox).
While I'm sure that it's possible to use modern techniques to estimate where crime will occur, it won't work in practice. There are simply too many ways to bias the results (intentionally or not). I suspect giving police this kind of tool is simply a way to give legitimacy and cover to their bad behavior.
> including information about friendships, social media activity
COINTELPRO is a helluva drug.
> advocates say predictive policing can help improve police-community relations by focusing on the people most likely to become involved in violent crime.
That sounds suspiciously like an excuse to improve white communities, by focusing on the blacks (who have historically been seen as "violent savages" by racists).
> because our predictive tool shows us you might commit a crime at some point in the future
The big question is how long until someone tries to use that "predictive tool" as probable cause.
To say that other tools have a bad track record may count as a valid argument, but to me, it is a weak and fatalistic one. Judge each tool on its own merit or discard all tools as useless seems like an easy choice.
Predicting crime works in practice and theory. These models are not black boxes, they can be introspected. Bias can be detected and removed.
COINTELPRO was a program to infiltrate and disturb organizations that the state viewed as unwelcome. Monitoring social media activity is common detective work. The modern equivalent of an officer peeking over the fence in your back garden to see a stolen motorbike. Now they can use Google Maps for that. This is public information: The criminals feel free and safe enough to post and brag about their crimes on Facebook.
Removing or combating criminal elements in any community will improve that community, regardless of skin color. Black youth is helped, not suppressed, when gang recruiters are identified and punished.
Predictive tools are already used as probable cause. Prisoners in Guantanamo Bay can get a brain-wave reader test. This device will tell you what someone is thinking about and may reveal the plans of future terrorist attack.
Yes. I know that. Apophenia is the correct term. Your fancy machine that predicts crime is only as good as the data it is fed and is made worse by the person who interprets the results. Both of these are easily biased.
> Bias can be detected and removed.
Just like how they removed the numerous biases, assumptions, and bad methodology that are well-known problems with breathalysers? Even if the model was theoretically accurate, the implementation can (and will) be wrong. You seem to be using a just-world assumption that doesn't have lazy, incompetent or malicious people.
> Monitoring social media is common detective work
It can be both. While these methods may be useful for going after stupid criminals, you're ignoring that it is also useful when targeting activists, political dissidents, etc. If you think this doesn't happen, you haven't been paying attention.
You're problem is that you are assuming it is only "criminals" that are targeted, but you live in a world where, to use an obvious example, some people assume that any black person is a "criminal".
Indeed. But remedies exist. Statisticians can examine the validity of the data, analysts and detectives can be trained to interpret the results correctly, and social scientists can point out the dangers of relying solely on computer systems.
I have no strong view for or against breathalysers. I'll concede that there may be some errors in those tests. Does that completely invalidate these tools in judging if someone is too drunk to drive and may cause harm to self or others? Should we only opt for rigorous methods like drawing blood samples? My view is: no, we should not. These are valuable tools that work for the large majority of times and help save lives (at the inevitable cost of some errors and inconvenience).
In my world I believe in just intentions. Breathalysers are not introduced to imprison sober drivers, they are to combat drunk ("lazy, incompetent or malicious") drivers on our roads.
These methods are useful for catching the savvy criminals too. I am not ignoring that these systems are also useful to target activists and political dissidents. That's basically what they were build for in the first place (well, that and the terrorists, see DARPA LifeLog). It's just now that these tools are adopted by local Police departments.
A weapon stick can be used to subdue a suspect through non-lethal force, and it can be used to choke a peaceful protester. It will succeed in both tasks. It's not a stupid ineffective tool we should take away, because it can be used in bad ways. We should make sure to avoid the bad usage, and provide police with the best weapon stick possible for the good usage.
You assume that this system will be used to justify police brutality and that this system will be used by people who think that any black person is a "criminal". I have a higher opinion of the people who join the police. I rather reserve such judgment to the criminals themselves.
Yes, when the error is this large and so easily manipulatable. (breathalysers are notorious for being incredibly broad in what the detect (bad false positive rate), and they are required to assume a 2100:1 ratio when estimating blood concentration from the measured breath concentration. In reality, there is a lot (up to +/-800 for some people). There is a good, science-based reason for that ratio involving the partial pressure of EtOH. The reason is valid, it simply ignores the (large) minority of cases where other factors complicate the analysis.
You may suggest that it would be easy to use modern techniques to find a better formula that accounts for these variations. That would work... but it has always been possible. You don't need anything particularly fancy to add a few corrections. These problems - and how to correct them - have been known for decades, yet breathalysers haven't changed. Why? Because an inaccurate tool gives police the leeway to target a much larger set of people (if they want to - selective enforcement is a powerful tool).
> Should we only opt for rigorous methods like drawing blood samples?
Yes, absolutely, and I (and many defence lawyers and civil right organizations) recommend insisting on a blood test should you ever asked to take a breathalyser because of how inaccurate and manipulatable the breath test is in real-world situations. (disclaimer: this can vary between states; see a local lawyer for proper advice)
> at the inevitable cost of some errors and inconvenience
A necessary feature of a free society is the assumption of innocence until proven guilty. Law enforcement is deliberately given a harder task, because errors are not simply an inconvenience. Errors risk charging an innocent person with a crime. You would not call being arrested because of a false-positive an inconvenience.
> You assume that this system will be used to justify police brutality and that this system will be used by people who think that any black person is a "criminal".
I don't need to assume anything. These things are already extremely common today without the need for advanced data processing techniques. If for some reason you doubt these facts, you may want to look up the per-capita incarceration rates by race and compare that to stuff like the drug use rates for the same groups.
> In my world I believe in just intentions
https://en.wikipedia.org/wiki/Just-world_fallacy
> I have a higher opinion of the people who join the police
I prefer to keep my opinions based in reality.
While I'm sure only a minority of police are abusive, the rest are aiding and abetting by not reporting the crimes committed other officers. Misprision of felony is still a crime in the US (18 U.S. Code § 4) ( https://www.law.cornell.edu/uscode/text/18/4 )
Misprision requires concealment and not reporting; not reporting by itself is insufficient, concealment is active. See, United States v. Johnson, 546 F.2d 1225 (5th Cir. 1977) [0]
[0] http://openjurist.org/546/f2d/1225/united-states-v-johnson
There is no shortage of police that are giving cover for the murders in their department, and most of the rest still look the other way.
So, yet again software is the new force-multiplier? Strict (IMNHO crazy) sentencing guides, arguably designed for the purpose of reducing crime by being a deterrent, now leads to even more filling up of prisons due to targeted (ab)use against certain groups?
On a side note, I wonder how these algorithms handle police brutality etc. I can just imagine sitting in such a meeting, and seeing a couple of police officers in full uniform popping up on that mugshot wall of shame...
Instead, it developed into a story of what amounts to pre-meditated blind rage against any and all associated with a given criminal.
This isn't new. And it's exactly what we don't need more of.
Step 2: Run the results of that through the computer.
Step 3: The unbiased computer tells me that disliked groups are more likely to be charged with crimes! That justifies my discrimination! I knew they were up to no good! And it's not me, the COMPUTER says it!
hoses self down with Ubik
Don't let the cartoon characters fool you. IMO The complexity and depth of this series (and a lot of sci-fi anime) far surpasses the likes of minority report and hollywood sci-fi in general. My description really doesn't do it justice.
“I was disturbed,” said Mr. Brown
Sounds like intimidation.
Interesting (eerie) parallel to the intro in the TV series "Person of Interest", although supposedly this system doesn't get its data from the NSA (as with parallel construction etc), but rather the information comes from more-or-less open data (legal surveillance etc) -- and of course it isn't vigilantes but police that will, 'victim or perpetrator, if your number's up (...) find you':
"You are being watched. The government has a secret system: a machine that spies on you every hour of every day. I know, because I built it. I designed the machine to detect acts of terror, but it sees everything. Violent crimes involving ordinary people; people like you. Crimes the government considered 'irrelevant'. They wouldn't act, so I decided I would. But I needed a partner, someone with the skills to intervene. Hunted by the authorities, we work in secret. You'll never find us, but victim or perpetrator, if your number's up... we'll find you".
Not to mention, I'm sure you can see how this is a breach of justice.
We can't outsource our moral imperative to computers, no matter how much we don't want to have the conversation about actual causes of crime and incarceration in the US...
An automated justice system wouldn't be biased by human prejudice, ignorance or fear. It doesn't get tired, doesn't feel pain or pity or remorse. It would be able to impartially and accurately process vast amounts of data - far more than a human, and the actions of the drones could be completely auditable. Drones won't lie on the stand to protect their fellow drones, or tamper with evidence.
You could walk down the street surrounded by police drones and be confident that you're not being profiled based on racist or religious bigotry, but pure mathematics and statistics. In every conceivable way, an armed drone with a license to kill is safer, faster, more reliable than a human. One only has to look at the current justice system in any country to see that humans are simply not capable of properly judging the motives of, or punishing, other humans in any reasonable way.
People will simply have to accept that the day will come when human-applied justice is viewed with the same ridicule and scorn that witch-burning is today... as the vicious, superstitious barbarism of an ignorant past.
An automated system is written by people and those prejudices can still sneak in...
http://www.nytimes.com/2015/07/10/upshot/when-algorithms-dis...
http://www.salon.com/2013/02/04/online_advertisings_racism_m...
Of course, being in the NYTimes and Salon, they need to obfuscate this point and appeal to standard mood affiliation.
Also, describing the results of any code/program written by a human as a "completely inhuman intelligence" is a tenuous claim at best.
If you want to make normative arguments, go ahead. My first principles tend to be very individualistic (I view individual humans as being the sole carriers of moral consideration), so our normative claims will likely disagree.
Also, describing the results of any code/program written by a human as a "completely inhuman intelligence" is a tenuous claim at best.
Clearly you've never written such systems. If they behaved remotely the way humans think my job (building them and making them usable by humans) would be vastly easier.
Appeals to authority and accomplishments aside, I don't need to have written such systems to understand, infer, and conclude things about aspects of their behavior. My point is this: something created by humans cannot be, by definition, inhuman. Two methodologies, the "human approach" and the "ML approach", might have radically different steps but come to the same conclusions. It would appear from your comments that you are OK with these conclusions ("An unbiased methodology produced these results, therefore, it's OK!"). Are you morally satisfied by the conclusions discussed above? Do the results of "such systems" influence your satisfaction?
The implication for society (assuming these findings generalize) is that most likely, we cannot solve statistical disparities via unbiased processes - we can have fair treatment of individuals or statistically representative distribution of spoils, but not both. As noted above, I'm very individualistic, so I favor fair treatment of individual humans.
Appeals to authority and accomplishments aside, I don't need to have written such systems to understand, infer, and conclude things about aspects of their behavior. My point is this: something created by humans cannot be, by definition, inhuman.
I don't know what you mean by "inhuman". It sounds like you mean the term to be "never tainted by the ephemeral emanations of humanity". I merely mean "inhuman" as "thought processes so radically different that intuitions about a human mind are completely useless".
Concretely, do you believe a random forest can somehow infer that the variable x[27] represents gender, and that to make it's sexist creator happy it should reduce the objective function in order to screw some women over? If you look at the internals of sklearn, that's just not what random forests do.
Two methodologies, the "human approach" and the "ML approach", might have radically different steps but come to the same conclusions. It would appear from your comments that you are OK with these conclusions ("An unbiased methodology produced these results, therefore, it's OK!"). Are you morally satisfied by the conclusions discussed above? Do the results of "such systems" influence your satisfaction?
I don't know what you mean by "morally satisfied". A fact about the world is either true or false. In computer science terms, I believe "morally satisfied" has type `satisfied: HumanAction -> Boolean`. Your question consists of applying `satisfied` to a value of type `WorldState` - it's a type error. In human terms, your question doesn't make sense.
In terms of my own individual happiness (as distinguished from moral satisfaction), this fact reduces my happiness. Because I believe many of these facts to be true, I'm forced to either lie about my beliefs (which causes me disutility) or suffer social opprobrium from anti-intellectual types and lazy thinkers influenced by them.
Even an unbiased analysis system can reach bad conclusions from bad data, and a biased justice system can produce bad data. So the conclusions can be biased even if the program is unbiased.
Is it your belief that we can cook up better algos/data collection methods/etc and all the people complaining about "bias" in algorithms will be satisfied? I don't believe that is the case, given that no one is actually complaining that the algorithms are getting the (factually) wrong answer.
If the arresting officers are biased, the number of arrests are probably biased. I'm not sure that you can fix that with better data collection methods; the problem isn't with the data collection. The problem also isn't with the analysis.
It may be possible, at least in theory, to create an algorithm that will determine whether the officers are biased, or whether the race in question actually commits more crimes. I'm not optimistic about that working in reality, though.
The core question - do you believe the problem will be fixed by better machine learning algorithms? Going back to the current example, do you believe that a Bayes-optimal machine learning algorithm for predicting criminal behavior will be "unbiased" (in the sense of social justice, not in the sense of statistics)?
Or, more concretely, do you believe that the only problem that mtgx and smtddr are complaining about is that our ML algorithms aren't good enough and that maybe we need deep learning instead of random forests?
When we get super-human AI, then maybe we can hope for that.