Santa Cruz, California bans predictive policing in U.S. first • Hacker News Board

Re: Santa Cruz, California bans predictive policing in U.S. first

#451

Post by HenryBemis » Sat, Jun 27, 2020, 1:33 AM UTC

momokoko wrote: post #397

cptskippy wrote: post #392

Earlier quoted context omitted.

> There's a worrying trend in the conversation these days that has shifted from "there's a serious problem with racism in how police go about their job" to "we need less police". The conversations aren't about less police, they're about defunding the police and putting that money to better use. If the police choose to reduce their numbers in order to afford riot gear, tear gas guns, armored personnel carriers, warrio…

> Throwing police at the problem of gang violence hasn't solved it. It hasn't made the neighborhoods safer. That is not true. We have solid data from New York during the Giuliani era and more recently Camden, NJ. Whether or not we agree with how they accomplished that is another matter. For example in New York a massive increase in arrests for nonviolent crime and much more proactively policing in order to check for…

People keep referring to the NY Giuliani era, but I think it's time to read Freakonomics and discover the true reason of the crime decreasing and it was definitely NOT Guiiliani's master-skills in public administration. Whatever grace he ever had, he lost it when he showed the world his true colours recently.

Re: Santa Cruz, California bans predictive policing in U.S. first

#452

Post by Misdicorl » Sat, Jun 27, 2020, 1:35 AM UTC

tick_tock_tick wrote: post #417

Misdicorl wrote: post #393

Earlier quoted context omitted.

Or we could address the root problem behind the vast majority of crime; systemic poverty.

I'm not convinced that systemic poverty is the cause. Plenty of poorer counties and non urban, but just a poor, areas in the USA don't have these high violent crime rates.

Poverty includes more than just household income. In particular, many of the comparative areas you mention provide: safe schools, libraries, extended family resources. Population density is also a huge driver of violent crime rates.

Of course poverty isn't the only factor. But it's a hugely important one.

Re: Santa Cruz, California bans predictive policing in U.S. first

#453

Post by lisper » Sat, Jun 27, 2020, 1:42 AM UTC

paragraft wrote: post #437

NickNameNick wrote: post #430

Earlier quoted context omitted.

Could you expand on that? I'm really curious about what the differences were.

One starting observation I'd venture is those countries (and Canada) base their policing on Peelian principles, which establish a basis of policing and the relationship with the public. The US police culture had different roots. Edit: for reference, the list of principles that were issued to each new police officer in London from 1829 onward. 1. The basic mission for which the police exist is to prevent crime and dis…

The problem is that these principles work great in a homogeneous society where everyone has some kind of stake in the societal outcome. If you take people with a particular skin color and enslave them and otherwise systematically discriminate against them to varying degrees over the course of 400 years then these principles break down. It is going to take a long, long time -- probably multiple generations -- for American society to recover from 400 years of institutionalized racism.

Re: Santa Cruz, California bans predictive policing in U.S. first

#454

Post by baby » Sat, Jun 27, 2020, 1:44 AM UTC

I’ve thought about something. I’m pretty sure that being single can really do some damages to you, or having a poor family more easily leads to you becoming a criminal. I’ve once heard someone saying that if everybody could get hooked then there would be no terrorism. Does it mean we should invest in surveillance and monitoring of people who are poor and single? Or should we instead fund aids, universal healthcare, free education or even UBI.

Re: Santa Cruz, California bans predictive policing in U.S. first

#455

Post by mikemotherwell » Sat, Jun 27, 2020, 1:45 AM UTC

snowwrestler wrote: post #306

alfalfasprout wrote: post #232

Earlier quoted context omitted.

I agree that, in general, you're going to have biased training data (with a bias that's difficult to measure) and so inherently policing recommendations will be biased. This is particularly problematic when it comes to arrests for crimes that are inherently 'selective' in their enforcement. Eg; drug-related crimes, public intoxication, loitering, trespassing. But the fact is... violent (fatal and nonfatal) crimes do…

If we want to stop neighborhood gang violence in the long term, we need to work on why people join gangs, and what they are fighting over. We can't do that with police, who generally don't come into the situation until after it's too late and things have gone wrong. But if most of our money is going to the police, then what resources are available to do the longer-term work on gangs and neighborhoods? In software dev…

What I think society needs is an rspec of government spending (I'm using rspec as it is clear, or should be, what that implies). Test driven government, in essence.

Society is varied across county, city, state and country lines, and we are in our infancy in understanding what policies work, when, where and why. Combine that with no overt expected outcome, i.e. do we expect this policy to have an absolute outcome (assert(absolute_wealth > X)) or if we want a comparative outcome (assert(relative_income > 0.77)), and we end up with arguments about spending that are judged by rhetoric rather than success or failure on what should be a knowable scale.

Being more overt in how society judges government spending, especially when the goal is a social good, gives us the best shot at spending more on effective interventions, and less on ineffective ones. I think it also helps mitigate the circus nature of politics, where politicians gain kudos simply for with their spending decisions, and are rarely held accountable to actual outcomes. Like TDD, it would by no means be a panacea, but it would be a step in the right direction.

While it true that a policy can be fantastic for society, but exacerbate inequalities (i.e. get better at extracted value from smart people), or be terrible for society but reduce inequality (just make sure everyone has nothing and job done), if we at least tested the policy along some overt lines, and threw out the very worst policies, I can't imagine social spending would have worse outcomes, if only until the loopholes are worked out.

Re: Santa Cruz, California bans predictive policing in U.S. first

#456

Post by blackrock » Sat, Jun 27, 2020, 2:10 AM UTC

Honestly, for all the money that they spend on cops and these high tech programs, there is a much simpler solution.

Give the poor people money!

Give them a safe place to live. Pay for their rent, or offset it with some rental assistance. Give them jobs. Give their children education, so that they can grow up to get better jobs. Give them medical care, so that they can stay healthy. Give them access to healthy foods so that they can eat healthy.

Stop arresting people for stupid things. And dismantle the Prison Industrial Governing System.

Have a heart already.

It’s really not a difficult problem to solve.

This is not a problem that requires advanced technology. It is a societal problem, that requires all of society to solve it.

America needs to spread the wealth. America needs to be more inclusive. Not exclusive!

Re: Santa Cruz, California bans predictive policing in U.S. first

#457

Post by monksy » Sat, Jun 27, 2020, 2:13 AM UTC

snowwrestler wrote: post #306

alfalfasprout wrote: post #232

Earlier quoted context omitted.

I agree that, in general, you're going to have biased training data (with a bias that's difficult to measure) and so inherently policing recommendations will be biased. This is particularly problematic when it comes to arrests for crimes that are inherently 'selective' in their enforcement. Eg; drug-related crimes, public intoxication, loitering, trespassing. But the fact is... violent (fatal and nonfatal) crimes do…

If we want to stop neighborhood gang violence in the long term, we need to work on why people join gangs, and what they are fighting over. We can't do that with police, who generally don't come into the situation until after it's too late and things have gone wrong. But if most of our money is going to the police, then what resources are available to do the longer-term work on gangs and neighborhoods? In software dev…

One of the reasons is: They can't trust the police to come out and take the people who cause trouble out.

Neighbourhood gangs, esp in chicago are territorial and they try to "protect the neighbourhoods." (I think they do cause more problems) This came up during the riots earlier this month. The police were stretched way too thin. Gangs did stop the looters and rioters from extending into their neighbourhoods. (Deeper within Logan Square and Little Village were an example)

Re: Santa Cruz, California bans predictive policing in U.S. first

#458

Post by danShumway » Sat, Jun 27, 2020, 2:14 AM UTC

treis wrote: post #156

basch wrote: post #111

Earlier quoted context omitted.

I like to bring Then Now up everytime precrime comes up. Until it goes off the rails in the end, its one of my favorite articles of all time. The paragraphs I yanked below, dont do its craziness justice. https://www.bbc.co.uk/blogs/adamcurtis/entries/78691781-c9b7... "But the oddest is STATIC-99. It's a way of predicting whether sex offenders are likely to commit crimes again after they have been released. In America…

The options here are having the judge make the decision based on their gut feeling or they can make it based on data. It's really hard for me to believe that gut feeling is better than data.

> It's really hard for me to believe that gut feeling is better than data.

This is exactly the problem. People assume that data means something by virtue of it being data.

Having a judge make a decision based on gut feelings might not be any more accurate, but at least people wouldn't assume that it was accurate. People would look at it and realize that it was arbitrary. If the data isn't giving good results, then at least we should use a system that makes that fact obvious.

> A recent very critical report of such systems said that the margin of error for individuals could be as great as between 5% and 95%

At a certain point, it would be just as accurate, and far more honest, for us to just flip a few coins. Get it out in the open what we're doing. Get rid of the built-in assumptions people have that the system is fair or predictive just because it's based on data. Make observers live with the fact that it's arbitrary.

Re: Santa Cruz, California bans predictive policing in U.S. first

#459

Post by datavirtue » Sat, Jun 27, 2020, 2:15 AM UTC

snowwrestler wrote: post #306

alfalfasprout wrote: post #232

Earlier quoted context omitted.

I agree that, in general, you're going to have biased training data (with a bias that's difficult to measure) and so inherently policing recommendations will be biased. This is particularly problematic when it comes to arrests for crimes that are inherently 'selective' in their enforcement. Eg; drug-related crimes, public intoxication, loitering, trespassing. But the fact is... violent (fatal and nonfatal) crimes do…

If we want to stop neighborhood gang violence in the long term, we need to work on why people join gangs, and what they are fighting over. We can't do that with police, who generally don't come into the situation until after it's too late and things have gone wrong. But if most of our money is going to the police, then what resources are available to do the longer-term work on gangs and neighborhoods? In software dev…

A wise manager would know that investing in better product development is an inglorious losing battle that is of little interest to people other than themselves (on an academic level) and to junior developers who simply don't know any better, yet. Your thought is adorable though.

Re: Santa Cruz, California bans predictive policing in U.S. first

#460

Post by monksy » Sat, Jun 27, 2020, 2:15 AM UTC

threatofrain wrote: post #411

alfalfasprout wrote: post #232

Earlier quoted context omitted.

I agree that, in general, you're going to have biased training data (with a bias that's difficult to measure) and so inherently policing recommendations will be biased. This is particularly problematic when it comes to arrests for crimes that are inherently 'selective' in their enforcement. Eg; drug-related crimes, public intoxication, loitering, trespassing. But the fact is... violent (fatal and nonfatal) crimes do…

Isn’t Chicago also where they discovered a police black site that had been operating for years? What’s the story here? https://www.theguardian.com/us-news/2015/feb/24/chicago-poli...

It got shut down. It sounds bad but you have to remember how violent some of the areas get on the West and South side (Garflied Park W+E, and Engelwood are all within the top 5)