Math That Gets You Arrested
Вставка
- Опубліковано 13 лис 2022
- One of the most significant developments in the history of policing is the use of statistics to track crime patterns and to determine how to react to them. New York City’s CompStat program has served as a model not just for cities around the United States, but also globally. And it makes sense, right? The better we track crimes and the more data we have, the more effectively we can allocate resources to improve public safety.
Unfortunately, New York City (and everyone else) has found that it isn’t that simple. Jack Maple’s bold vision for a statistics-based police department has been plagued by inconsistent application and perverse incentives that prioritize numbers over public safety. The perpetual conflict between good policing and good CompStat numbers has mitigated the program’s positive effects and magnified its criticism on civil rights grounds.
CompStat reinforces our biggest problem with statistics in public life: the numbers we increasingly depend on don’t lie, but we don’t always know which truth they’re telling us.
** ADDITIONAL READING **
Jack Maple, “The Crime Fighter: Putting the Bad Guys Out of Business” (1999) www.amazon.com/Crime-Fighter-...
“Compstat: Its Origins, Evolution, and Future in Law Enforcement Agencies,” US Department of Justice, Office of Justice Programs: bja.ojp.gov/sites/g/files/xyc...
George Kelling, “Fixing Broken Windows: Restoring Order And Reducing Crime In Our Communities” www.amazon.com/Fixing-Broken-...
NYPD’s CompStat 2.0: compstat.nypdonline.org/2e5c3...
** LINKS **
Vsauce2:
TikTok: / vsaucetwo
Twitter: / vsaucetwo
Facebook: / vsaucetwo
Talk Vsauce2 in The Create Unknown Discord: / discord
Vsauce2 on Reddit: / vsauce2
Hosted and Produced by Kevin Lieber
Instagram: / kevlieber
Twitter: / kevinlieber
Podcast: / thecreateunknown
Research and Writing by Matthew Tabor
/ tabortcu
Editing by John Swan
/ @johnswanyt
Huge Thanks To Paula Lieber
www.etsy.com/shop/Craftality
Vsauce's Curiosity Box: www.curiositybox.com/
#education #vsauce #crime - Наука та технологія
YOU HAVE ROUGHLY 24 YEARS OF SLEEPING TO DO. Do it on a great mattress. Visit helixsleep.com/vsaucetwo to get up to $200 off your Helix mattress, plus two free pillows #helixsleep
*24 years of trying to sleep
Gotta remember the insomnicas
Ž
Reminds me of a friend who was put on a sex offenders list. Reason: He was at a party, got drunk, decided not to drive home as that was wrong. So he walked home, but halfway home needed to pee. He found the first tree, let loose, and was quickly apprehended.
Initially he was charged with public indecency, but it was later upgraded because the tree he happened to pee on was on a public school ground.
The time of the arrest was 2 in the morning, but still he was labeled a sex offender. It took him a solid three years of lawyers and court debates before he was wiped from the list, and never once was he given an apology. Last time I spoke with him he was still being harassed by people because his name is on old lists. And this all occurred in 2010.
Apparently a cop wanted to boost the stats, and a guy is still dealing with the ramifications of that to this day. Why? Because he chose to walk home drunk rather than drive. No good deed goes unpunished.
Cops have done that and a lot worse too. Letting Serial Killers go free while shooting innocent dark-skinned people.
No good pee goes unpunished. I'm sorry I couldn't help it.
Meanwhile, that cop had at least a 40% chance of being a domestic abuser.
At least he didn't defecate through a sun roof
Exactly. This is such a great real world example of how over policing is not beneficial to public safety. Social safety nets are, however.
This makes me think of a problem over here in the UK when NHS dentists were paid per filling they did. As a result dentists started doing random fillings on anybody if they needed them or not while disincentivising promoting preventative dental care meaning everyone's teeth got worse and dentists mad bank.
that's Keynes's practice of hire people to dig holes and other people to fill them.
That's what happened to me when I was a child... they drilled out my perfectly good molar tooth and filled it. Never had any cavities or fillings ever since, and I'm almost 50 now...
Every dentist I have ever gone to.
Wow is that true? I thought it was an urban legend!
Reminds me of when the Indian government put a bounty on snakes to try and keep the population under control
So people started breeding snakes to claim the bounty
But then the Indian government got wise to it and stopped paying out bounties... so the breeders just released all of their worthless snakes and the problem was worse than before
This is a good example of Goodhart's law "When a measure becomes a target, it ceases to be a good measure" and can be applied to so many things in our lives, some examples I've seen listed here is the UK having paid dentists per filling they did, meaning that they recommended fillings even for people who didn't need it, and wouldn't promote preventative dental care. Another is education and testing, the measure of how much you know is a test, so now the goal is not to have knowledge but to do well in the test.
This is also a very common problem in AI, if you give an AI 1 point for each step it takes in a maze, and 100 for reaching the goal, it might just start spinning in place if it thinks reaching the end of the maze is too hard
Confusing the map for the territory
“Crime will never be zero”. A profound statement amidst a video about policing.
The fact that police have a "quota" is the most backwards thing I've ever heard.
To me, it sounds like any business that's gotten big enough to be called political / adopt "performance metrics" for shareholders.
It's weird because I worked for my university's police and safety department as a meter maid and was responsible for writing parking tickets. We had an unofficial quota of 6-10 tickets per hour. It was weird because I felt like I had to hunt down the smallest details of any parking violation to keep my job. I left after two years.
there's a famous quote "when a measure becomes a target, it fails to be a good measure". which I feel is the exact state compstat is in at the moment.
My ex was loud when she didn't get her way so after the cops were called multiple times, they arrested me, the man because they said the man is usually the aggressor. We soon broke up and I have never dealt with the cops since, she on the other hand went to jail 6 months after for beating up an elderly woman and her mentally disabled son with a brick in her purse. Stats have variables
@1 2 , I wouldn't consider this an outlier at all. Even if it is more common for the man to be the aggressor in domestic disputes, the number of times it is the woman that's the aggressor is not so low as to be called an outlier (given that it's two options, it only has to be the man 50.1% to the woman's 49.9% for it to be more commonly the man). Besides, this shows the statistic itself is unreliable. His case would reflect in the statistic as the man being the one arrested, but that became part of the statistic entirely because of the statistic. It's a feedback loop.
@@SgtSupaman Women are actually considerably more likely than men to be the sole aggressor in heterosexual relationships, they just don't get the cops called on them most of the time. That said, sole aggressors are rare, it's usually both of them that are violent.
This reminds me of an old saying in Economics: “show me the incentives, I’ll tell you the outcome.”
Hey Kevin! Police officer for a small town here! Love the video! Luckily, my department doesn’t use our data in such a way as “we need to increase ‘x’ or decrease’y’. We obviously still use the data, but more in the sense of, there’s been three car break ins on Main St., time to patrol Main St. more often throughout the shift. We respond to the calls we get and we patrol and interact with our community. I know not all places can say the same (obviously, that’s what the video was about). But I still appreciate you doing the research and making this video! Keep up the excellent work
Doesn't change the fact that when you're patrolling Main st you're more likely to stop and question a brown person and that's why ACAB
Good job!
It's the old programmer's adage: garbage in, garbage out. When you don't properly control the input (in this case, the way crime stats are collected), then the process will usually spit out corrupted data as the result.
Maybe the real garbage was the malicious police officers we made along the way.
@@vigilantcosmicpenguin8721 lol. funny and very accurate
thats exactly how we tune engines. If a sensor isnt calibrated right or the software isnt matching what the engine is doing. It could be the difference between a blown up engine or a engine that can win races. Ive personally have synced the software to the engine about 10degrees off from the rotation of the engine and that caused a serious failure.
Same goes with audio and video, can't always just fix it in post.
This is another one of those videos that makes me think back to my GCSE psychology teacher, he said "statistics are like a bikini. What they reveal is enticing, but what they conceal is vital" I love statistics think they can be amazing but only when I know I have all the data and know it is accurate
there are 3 types of lies, lies, damn lies and statistics
This video is not about inaccurate data.
Yeah I know, but it is about using statistics in a way you want to, to show what you want. So the statistics are not entirely accurate, as was mentioned in the video what is easily measured. So we know the statistics used are not the full picture. The accuracy part was just part of my opinion on when I do like to use statistics
@@Rayzan1000 , it partially is, though it is just one aspect of why there is an issue. As stated in the video, you get more of what you measure. Statistics become skewed because of what is being focused on. For instance, if I see there is more crime occurring in neighborhood A than neighborhood B, I'll send more units to patrol neighborhood A. More patrols means they will notice more crimes. Now neighborhood A's crime rate is actually going up while B's is staying the same (even though there might be more crime happening in B that just isn't getting caught).
When a metric becomes a target, it ceases to be a good metric. I think I heard that in Robert Miles's videos about AI alignment.
If a metric becoming a target makes it no longer a good metric, than it was never a good metric
I recall one of the Adam Curtis documentaries about how governments turned to objective data gathering to set economic policy, which set the stage for the numerous scandals.
Seen the same in healthcare, where fragile patients become a hot potato to be on someone else's mortality chart.
@@ProductBasement Untrue. There are a great many metrics that have historically made perfect proxies for whatever people are actually interested in, right up until the people being measured heard about them being used as proxies - at which point they started optimising the proxies, not the desired outcome(s). They're perfectly fine as metrics so long as they're secret. The problem is that unless you can manage to find a way to measure whatever it is you're actually interested in (which is, in general, a very hard problem) you have to use something as a proxy, and it's another very hard problem to find proxies that can't produce perverse incentives once they become known targets.
It's amazing how humans will always find the wrong way to use something. FFS HOW can solving a burglary not have more significance than issuing tickets?
Tickets result in revenue for the government. Solving a burglary just protects the public, who could protect themselves with security systems that increase revenue in the local economy and thus revenue for government...
@@michaelrichter9427 So they can't even measure their own measurements properly.
Especially as jaywalking is the ultimate victimless crime, in fact the perpetrator is the one most in danger of a hit and run
It's said in the vid: it's based on (still controversial) "Broken Windows" theory.
Since it's not very well explained: the theory itself goes like this:
"If you have a street where all the windows are broken, graffiti on all the houses etc; it tells you visually that small crime is prevalent in the area. If you know small crime is prevalent; it will make it easier to start committing medium and high level crimes.
Therefore: if you focus all your attention on removing the vandals; meaning all the windows and walls stay intact and clean, it will disincentivise people from wanting to do medium crimes, since everything looks more secure."
As mentioned; the theory itself is very controversial, but these people ran with it full force. It's just been expanded from mere property crime to "any crime that is easy to see the results of".
Because the data says so
Would love to see a VS2 video about how the same thing essentially applies to school's slapping standardized testing on our nations youth.
I was wondering the same thing!
Yes!!!
I almost feel like this episode ended too early.
I feel as if I'm left with a bigger hole to the questions. Like I'm trying to dig around me to find the answer, but in reality all I've done is dig myself deeper.
And that's kind of what this video feels like, no solution given.. just the facts.
I think that is the point. A 10/15 minute video can not solve crime, where professionals have failed using and refined a method that "back in the days" was literally better then an annual report.
They had to end the video because the obvious conclusion is that acab and policing needs to be abolished, but that's too dramatic for a UA-cam video that has to be impartial for broad appeal
- I need a doctor!
- I'm a doctor in mathematics
- My friend is dying!!
- Minus one..
I’ve been subscribes for like a decade, but this is the first time you came up on my feed in about five years….
Same here
Bro same
same
Same 😮
Good old UA-cam algorithm! Hope you enjoy the new video and thanks for being a longtime sub.
Great video, these are so good for introducing people to the topic and demonstrating the problems, amazing work, much respect!
I have to say back when these videos were a brief bit of screaming about odd maths trivia I never would have guessed they'd evolve into powerful social commentary, not that i didn't love the old videos of course but I really love these.
"Juking the stats. If you juke the stats, majors become colonels. I've been here before." - Roland Pryzbylewski
Great video. It tells you about the benefits and dangers of technology. As someone who's been working in the technology industry for over 20 years it's a curse and a blessing....
I love that you've been putting your focus on things like this.
you're very quickly turning into one of my favourite channels out there!
the "cobra effect", I think I learned about it here: In India, British colonial authorities paid a bounty for every dead cobra. This led to an increase in breeding snakes for profit, and the eventual release of the snakes into the wild when they could no longer breed them at a sufficient rate
Wasn't because of the breed rate but because the bounty was remove when the authorities realized that there where people cheesing it
And keeping cobra to breed them was useless so ...
Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure.
When a measure becomes a target, it ceases to be a good measure.
Very good video! This reminds me of this very interesting chapter about the broken windows theory in the dutch book "de meeste mensen deugen" (idk what the translation is called exactly but something like 'most people are good')
It’s interesting how this is baked into our relationship with policing. People know to be careful when driving around at the end of the month because cops want to raise numbers and may be more likely to pull people over.
I never thought about it like this. So interesting it was designed in this way and never improved.
Very cool video, I hope the clear message is heard
Correlation does not equal causation. I think the real reason that crime rates went down is that leaded gasoline was banned in the 90's. In every country in which it's been measured, the crime rate perfectly tracks the lead levels in children's bones offset by 20 years. Veritasium had a video on the guy who invented leaded gasoline it's scary stuff.
Leaded gasoline is one huge factor but to say it is the *only* factor is wrong because crime isn’t a black and white thing
It’s a grey issue all around, there are an infinite reasons (valid or not) someone commits a crime, what gets classed as a crime and how the law is enforced all directly affect crime statistics in distinct ways and for the most part, we can track down certain trends like lead in gasoline or for this video, compstats
While leaded gas has influenced crime stats, so have policies like stop and frisk and police quotas that Kevin mentions
These videos are all so eye opening, and also heartbreaking.
Vsauce2 seems to have focussed in on the perverse interactions between the justice system and mathematics/information science . . . and I'm here for it!
Perverse incentives are everywhere when you try to oversimplify things into numbers
The only vsauce that uploads frequently.
3 things I have to mention about this video.
First of all, Kevin, you never disappoint. Always have loved and will forever love your videos.
Secondly, look at what we are doing us a society just because a small, but not negligible, portion of people decide to just be menaces to said society.
And lastly, and definitely my most important point, @5:48 the colour blending of the text was pretty cool, never seen something quite like it. Huh, you learn something every day.
Kinda looks like the inside of the text is just in inverted color, like the negative mode in some cameras
Funny he said “gotta catchem all” right as my Pokémon Scarlet download is finishing
I feel like good policing should be the goal of police departments and good compstat should be the job of social programs that can reduce the likelihood of someone being in a situation where they're more likely to commit crimes.
Yeah exactly. If your statistics show that the people in a certain neighbourhood are causing most of the crime, increasing the amount of policing in that area doesn't solve the issue. It only puts a bandaid on it. The true solution to crime is always to examine what is causing it and then try to action to eliminate those reasons.
Your videos are always great. Thanks dude
This, coupled with all of the other "problems" with policing nowadays makes me really think that the system we have at the moment needs overhauled.
This is why you don't just rely on percentages you also examine overall instances and then compare them to the rates. There should also CERTAINLY be a focus on the more serious violent crimes over any of these smaller offenses.
Good stuff. Didn't expect this kind of commentary
Kevin, thanks for keeping the Vsauce group alive!
I'm assuming that the people in charge of CompStat are well aware of the foibles you've pointed out. Would like to be a fly on the wall at their meetings and see the proposals and mechanisms they put forth to address these.
You really should do a companion piece with this in education
I think this is the first video about police I’ve watched in a year that didn’t talk about racism.
Maybe there's an issue with policing.
wow, it's almost like cops are racist, huh?
Tell me you sit in a far-left echo-chamber without saying it...
There is racism in policing, but I’ve seen a lot of normal policing that gets called racist just because. There is no clear answer.
@@meatlord-vq9bw No, there is not. As proven by FBI statistics. Stop lying to push a racist anti-white agenda.
10:29 I expected a "WRONG!" to follow
Another great piece on this exact topic is episodes 127&128 of the podcast Reply All.
oughtta be careful with this video before your local PD decides to harass you
"POLICING UNDER COMPSTAT IS A PERFECTLY BALANCED SYSTEM"
lol, would be a great video, but I'ld have no clue about how he would be able to pull something like that off. ;)
Maybe the real good cop was the friends we made along the way
I'm quite pleased this issue is getting more mainstream attention.
Pretty much summarized Season 3 of The Wire in minutes
Nooooo don't google "40% police"
Everything getting paid in taxes is gamifying itself to leech more efficiently.
Gotta catch em all
You used the Pokémon analogy while I was playing Pokémon.
Pinkman has come a long way since that dude with cancer died. My dude is now teaching maths instead of making 'chemicals'.
Excellent video Kevin.
A very interesting video. Thank you and well done
Man this make me want to rewatch the wire
So it's the same problem as in the corporate world, the people who get promotions and become managers are often terrible managers because they weren't good at managing or people skills, they were good at meeting company targets even if that came at the cost of efficiency, quality, client satisfaction and high turnover.
I think it's one of the causes of crunches un video-games industries
With employees becoming managers and are forcing crunches because it was their old solution even if it's the worst
Numbers are a dangerous game, in any scenario. Most of the time numbers go up something usually falls through the cracks.
I think I remember someone said to me, Numbers don’t lie, but that doesn’t me people know know what they’re saying. It also like when little ones over do something because they heard doing something with make them happy, so they over doing it will make them hyper happy. I guess it isn’t just kids now that I think about it….
Compstat information should only be used im high level analysis and less used to influence officer day to day policing. They should be used to deploy but not tell officers what to look for just to do their job in a specific area
Goodhart's Law: any measurement that becomes a target in itself then ceases to be a good measurement.
Just don't do crime, people
Hearing the motorsport manager OST really threw me off in a good way
Another reason why plea deals should be illegal. Often times an officer/detective will give you a plea deal because collecting and gathering evidence is a long, time consuming, expensive process. If they can scare you into accepting a plea deal they don’t have to worry about any of that and it satisfy their quota. No trial, no evidence, just scare a confession out of them and move on without caring if you got the right person or not. If every case wen to a trial, the court system would get backed up, and it would force police to focus on the real crimes and not just “pad their stats”
Unfortunately, instead, we get bail reform which means we just let everyone out to continue to do the same thing they were just picked up for, hence why you get a ton of repeat offenders. It’s not so much that crime is up, it’s that the criminals are being let back out into the streets.
Courts give out plea deals, not officers or investigators....
@@D35TR0YM4N Yes but hating cops is popular among teenagers so they'll complain about them whenever they can.
That sir, was well done.
compstat has one major flaw.
The morals of man isn't something that's properly measurable, it's random.
And if there's one thing we know about the properties of random numbers, it's that over time, over a large number of dice rolls, the frequency of each number showing becomes equal.
This is the same issue I take with prosecuters who tout a 100% conviction rate. Like you never went against someone who was innocent? If you did you put them away. What kind of games did you play to put an innocent person away?
Vsauce! Kevin here -- with 40%...
Another banger. Thanks!
It's irritating that no actual data was shown about whether any of these policies corresponded with a growth or decrease in violent/nonviolent crimes, I would have liked to have seen such charts and numbers.
CompStat is an absolute nightmare.
Rudy Giuliani is the American Royal, you know.. closely bred, red and needy.
There's one massive flaw with Compstat: it's reactive.
If you only react to something, you will never be able to control it and/or prevent it.
That's not Compstats fault though: it's just a database. It was never designed to be a program that predicts or controls anything: it's a catalogue for things that happened, it's just a Wikipedia for crimes.
A tax invoice for a business isn't there to tell you how to turn a profit; that's the job of the people running the buisness. Same thing here.
@@MikeMozzaro All of us in this comment section know that. But decision-making people that work with it day-to-day don't. :(
Looks like an example of Goodhart's law where the metric becomes the goal.
Can you do a similar video on how stats have changed baseball for the worse
You will always have people who "game the system".
its rough how Kevin does the math vids. Math gets no views but it the most interesting
I feel like he's enjoying this too much
Good old incomplete data biases. Great vid.
It it possible to put an ear on that outro theme? Has it the end and the beginning?
Damn thought this was gonna be ab counting cards at the casino or smthn
Policing ❌
Taking advantage of an old algorithm ✅
That 17% means the probability for Bara (Spirit Breaker) to bash a character in Dota 2 game. It rounds up to 100% if it's your opponent playing the hero, or to 0% if it's you playing the hero.
They did a documentary on the subject in the early 2000's. It's called The Wire.
I think you may have pinpointed why everyone hates cops
When is Kevin going to just make an episode called ACAB?
1:16 yo i have that exact calculator
Time to look up Robert Peel's policing principals.
They are coming up for their 200th anniversary so may be a bit dusty.
This same applies to "Equity Scores"
It has to be computer statistics. No one is spending the cash on a computer back then and not adding the word computer in there.
I: Great video, Kevin, but no cop is a "good cop". While individual police members might be good people, they are still cogs of a brutal, repressive system that does more harm than good and isn´t even that effective at combatting the kind of crime that leaves victims.
I sleep 7-9 hours a night, I'm in bed for 10. I've found my body knows BEFORE ME how much I need, and on rare occasion I'll sleep for all 10 hours.
I think a typical human gets way less and it shows, sleep is super important for keeping your edge.
what does 52/13 equal?
It is exactly equal to 104/26.
Interestingly enough, the most shocking thing in this video was the "New York's 98 police precincts". I guess the TV hast just been lying to me for years about there being a 99th one.
That's probably why they chose the 99th; to make sure the fictional precinct doesn't share a name with a real one.
9:08 John Stossel