345
Predictive Policing Software Is Terrible At Predicting Crimes
(themarkup.org)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 03 Oct 2023
345 points (97.8% liked)
Technology
59598 readers
2192 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
Forever.
I mean that's just not a realistic thing to believe. People aren't actually unique or special.
So at some point computational power will meet the right algorithm and suddenly we can model morons.
Short sprint to predictive policing. And before anyone gets all bent out of shape, go ahead and ask a criminal defense attorney how many of their clients are 'criminally stupid.' Based on conversations I've had, I imagine the answer is 'a fuck ton of them.'
Of course that makes people feel weird so we probably won't do it even when we can, but that's not the same thing as 'it'll never be possible.'
There is no algorithm that will ever predict that someone will do something100% because there are too many factors, including those that come up during the opportunity to commit a crime, to account for. That doesn't even cover the fact that the algorithm can only predict based on the information it is given and calculate based on our assumptions about people based on other data.
At best it will be the technical equivalent of stop and frisk, with racist outcomes based on racist assumptions. Like most forensic stuff, it will just be technology used to justify what people already assume.
Not to mention that stupid people doing stupid things makes them very unpredictable at the individual level.
There is lot wrong with your comment.... And what you're saying makes me suspect you have zero actual experience with the criminal justice system.
Just one point before I duck out - we put folks in jail for life and even kill them on less than 100% proof because the standard is 'reasonable doubt.'
Speculate all you want, enjoy being wrong.
For example, it is beyond a reasonable doubt.
There is far too much randomness in life to be able to predict everything, unless you can know everyone's actions at all points in time. Which we seem to not be too far off from...
Highly unlikely that'll be the case forever. We can already do population level behavioral prediction for advertising purposes. It's just a matter of time, quality data generation, and finding the right algorithm before we will be able to accurately predict where and when police resources should be deployed to efficiently deter crime. Especially since we already have a decent idea as to the factors that generally lead to spikes in crime-rates things like: poverty, widespread social isolation and low social cohesion, alcohol and drug use, perceived opportunity, and the presence of easily victimized populations such as racial minorities, religious minorities, the disabled, and the LGBT+ community.
Tbh, we don't even need such an algorithm because we already know that the best ways to reduce crime are to increase protections for those minorities, alleviate poverty, reduce the presence of alcohol selling establishments, provide addiction/mental illness care, promote social cohesion, and have community events where law enforcement builds trust and bonds with their local communities, promoting co-operation and mutual respect between law enforcement and the people they are supposed to protect. In other words, the best ways to combat crime are the exact opposite of what everyone in the USA has generally been doing, especially conservative areas. Predictive policing is only even desirable because we don't want to do the hard work of actually improving people's lives and building communities where crime isn't something people have/want to consider.
At best we can do what we already do and make estimates about groups, kind of like how we can have a fairly accurate predictions for climate but weather for a specific person's house is extremely unreliable in any detail a couple days ahead except for massive weather systems like hurricanes. As you noted we already know the causes, but trends do not predict which individuals will commit crimes.
There will be no point in time that an algorithm will be able to predict that an individual will commit a crime at a specific point in time.
I think we might've had a bit of miscommunication here. I wasn't talking about predictive policing at an individual level, that's highly unlikely to be possible, at least with traditional computing technologies (not to mention that individual predictive policing isn't even desirable for a multitude of reasons explored by many dystopian fiction authors throughout history) but rather at an area level. Being able to predict where and when crimes are likely to occur and with regularity, predicting that a specific drug store will probably be robbed within a narrow window of time for example. Even if such an algorithm was only accurate within a couple of hours it would fundamentally change how law enforcement functions, as well as the purpose it serves. Instead of merely enforcing the law after a crime is committed they could prevent crime/catch the criminal mid act without the need for informants, and without even knowing who they are gonna be arresting prior to catching them.