by parkaboy on 1/24/24, 3:35 PM with 91 comments
by Cheer2171 on 1/24/24, 4:45 PM
It is a basic statistical fallacy literally called the prosecutor's fallacy, because it gets misued so much in criminal law, also called the base rate fallacy. 99.9% accuracy sounds high, and is above what state of the art performance is for noisy real-world surveillance data.
But imagine this system is used on a city with a population of 1 million people. With an accuracy rate of 99.9%, the system will correctly identify a criminal as a criminal 99.9% of the time. But the actual proportion of criminals is, say, 0.1% or 1,000 criminals out of 1 million people.
When the system scans all 1 million people, it correctly identifies about 999 of the 1,000 criminals (99.9% of 1,000). But here's the catch: it also mistakenly identifies 0.1% of the 999,000 innocent people as criminals. That's 999 false positives. So, the system ends up flagging nearly 1,998 individuals (999 true criminals + 999 false positives) as criminals.
by saagarjha on 1/24/24, 3:49 PM
(We should also probably not rely on AI to put people in that place.)
by bevacqua on 1/24/24, 3:47 PM
AI didn't wrongly convict anyone. People leveraging AI did.
by bmitc on 1/24/24, 4:06 PM
> However, just hours before he was released from jail
How does a person not just walk out of the courtroom after having charges dismissed against them? This case is harrowing.
Also, there are multiple levels of failures here. Why is a person accused of robbery, who just had charges dismissed against them, in jail in the first place, and furthermore, why are they being housed with violent criminals? Why were they arrested purely on the basis that they look like someone, as determined by an automated system, with no further investigation or oversight?
by rbetts on 1/24/24, 4:02 PM
https://www.houstonpublicmedia.org/articles/technology/2024/...
by drdeca on 1/24/24, 3:55 PM
That’s probably not actually a sound policy, but, I think it is important to bring the average number of prison rapes per year to basically zero, very quickly, and this is the first thing I think of towards that goal.
... I suppose even with plenty of footage there may be the possibility of erring in determining whether it was consensual?
by enobrev on 1/24/24, 3:48 PM
by jMyles on 1/24/24, 4:02 PM
Putting pre-trial detainees in the same population as convicted prisoners (ie, county jails) is very strange.
by ashleyn on 1/24/24, 3:46 PM
by rhplus on 1/24/24, 3:52 PM
by ETHisso2017 on 1/24/24, 3:44 PM
by highwaylights on 1/24/24, 3:58 PM
by m3kw9 on 1/24/24, 4:04 PM
by fennecfoxy on 1/26/24, 11:59 AM
by hn_throwaway_99 on 1/24/24, 3:46 PM
But that being taken to jail in America can result in violent prison rape (and worse, that it is often met with jokes and snickers) shows a very sad, disgusting part of our society.
The article says he's suing EssilorLuxottica but I also hope he is able to sue the crap out of the government that incarcerated him.
by gigel82 on 1/24/24, 4:49 PM
by NoMoreNicksLeft on 1/24/24, 3:50 PM
by lifestyleguru on 1/24/24, 3:53 PM
by feverzsj on 1/24/24, 4:01 PM
by par on 1/24/24, 3:51 PM
by Teknomancer on 1/24/24, 5:04 PM
Why is nobody asking this question??
It seems like that should be a substantial part of the story.
by paxys on 1/24/24, 3:46 PM