Hacker News Reader

Comment by philipkglass | Hacker News Reader

I was somewhat surprised by how many negative comments there were on the Ars article.

I think it's important not to knee-jerk in favor of everything that could help to catch criminals. From my own post history on Ars (I use the same nym there) you can see that I am strongly against dragnet communications surveillance and strongly in favor of secure systems with no back doors for government access. But I think it's also important to not knee-jerk against every technology that can increase the effectiveness of law enforcement.

This particular approach seems on balance positive to me: it's targeted at catching already-known criminals. It's not invasive of the privacy of the general public. It would be a bad tool to have in the hands of a totalitarian dictatorship, but so would any effective law enforcement system.

View on HN

In this case it worked, but what are the false positive rate? Roughly 412 million images was scanned in this case. How many scans will this be per year if used on all known criminals that goes on the run? What happens in the case of a false positive, and at what rate will this cause major impact on an innocent (ie, lost time to fight the false accusation, lost jobs/relationships, swat team charging in with guns into someones home, and so on). Is there a cost-benefit analyze that is favorable to run this kind of biometric system over the more traditional hire-more-investigators approach to solving more cases? If it was cost-effective, would it still be so if the state provided a fair compensation to every harmed innocent?

Its not a knee-jerk reaction, but plain experience from reading the answer to such question that studies have done on past biometric, data retention and surveillance systems. A common theme that keep getting repeated are generally high cost, high false positive rates, and low results.

And how does one defend themselves after being falsely identified? I'm sure there's more that one guy that lived in Indiana 17 years ago that looked kinda similar to what that guy looked like then.

Having the FBI come into your workplace and arrest you is bad for your career whether you're guilty or not.

I remember when I was doing online dating -- I had the same name, had lived in the same area, and was about the same age as the brother of a serial killer that was quoted prominently in the news. I told new matches that when they googled me, that that guy was not me.

Not sure how to do convince my boss that when the FBI leads me out of the office in handcuffs that it's just a case of mistaken identity.

People have been identifying wanted criminals from photographs for a long time. Sometimes humans identify correctly but often they are incorrect.

I would hope that there is a less drastic/foolish procedure than "FBI handcuffs person at work, investigates later" every time that someone thinks their new neighbor looks like a criminal on a poster and phones in a tip. Just substitute "algorithm phones in a tip" for "neighbor phones in a tip."

https://www.usmarshals.gov/investigations/most_wanted/

High false positive rates are the one genuine concern, but requiring juries to be instructed about false positives (and ruling frankly misleading statements about probability of an individual match being correct inadmissible) and investigating for corroborating/confounding evidence seems like a more fruitful approach than resisting the technology.

It's not like other forms of police work don't involve many people inconvenienced by being incorrectly dragged into criminal cases and sometimes put on trial. I doubt the false positive rate from biometrics etc is worse than any other method of criminal investigation, and it's not to incorrectly-identified suspects' disadvantage that they're further removed from the crime than suspects incorrectly identified via more traditional investigative work.

For me, it's hard to speak against the outcome; criminals are being caught, after all. I'm wary, however, to give the government total leeway with technology. It's the issue of giving an inch to someone, and they take a mile: It (potentially) opens the door for misuse later on down the road. Whether or not a society is a democracy or a dictatorship is irrelevant, as either one can abuse the technology available to it.

I never have clear answers in cases like this one, because there is that balance between security and privacy. So far this seems to be good thing, but who knows what kind of abuse will happen down the road.

I wouldn't grant total leeway with other ways of using this technology either. I'd be upset if the government were doing the same biometric analysis on faces extracted from CCTV feeds across urban areas and archiving the results, for example. But processing photographs taken for government ID to see if they match wanted criminals seems pretty non-objectionable to me.

Reply on news.ycombinator.com