The citymeetings.nyc logo showing a pigeon at a podium with a microphone.

citymeetings.nyc

Your guide to NYC's public proceedings.

Q&A

False positive rates and addressing misidentification in facial recognition

1:14:38

ยท

153 sec

Council Member Yusef Salaam inquires about the false positive rate of NYPD's facial recognition system and how the department addresses cases of misidentification. NYPD officials respond with their processes and safeguards.

  • NYPD claims no wrongful convictions have resulted from their use of facial recognition technology
  • Multiple layers of human review are involved in the process
  • Facial recognition is used as an investigative tool, not as evidence at trial
  • NYPD conducts additional checks and investigations beyond the initial facial recognition match
  • Investigators look for physical characteristics and corroborative factors beyond just the facial match
Yusef Salaam
1:14:38
Just have a few more before I pass it.
1:14:40
What is the false positive rate of the NYPD's facial recognition system and how does the department address cases of misidentification?
Josh Levin
1:14:55
So as I understand that there has never once been a wrongful conviction in NYPD's use of this facial recognition technology in order to determine whether somebody is actually a suspect of interest.
1:15:08
So what I think the important thing to think about is there's multiple layers of human review on this.
1:15:13
You have to keep a human in the loop.
1:15:15
And what that does is allows a human being to look and compare.
1:15:19
And again, this is just to help us figure out the identity of somebody.
1:15:22
This is not evidence at trial.
1:15:24
You might have somebody using homophobic or, this is an actual example, slurs on a train.
1:15:29
And the person on the train doesn't know who this person is, but they snap a picture of them before they walk out.
1:15:35
Nobody knows who this person is.
1:15:37
And so this technology allows us to figure out an identity, which then we can do a further investigation.
1:15:42
And in regards to your specific concern, we do many checks internally once the machine gives us a possible match.
1:15:49
We have human beings who are checking what's that person's pedigree.
1:15:54
Do they even live in the city?
1:15:56
Do they have proof that they were outside of the city at the time this incident happened?
1:15:59
So just because it gives us that as a lead, we do all these extra things to try to build an outside case to determine could it even be this person?
1:16:07
Were they in a morgue, God forbid?
1:16:09
Were they in a hospital at the time?
1:16:10
Were they incarcerated somewhere else?
1:16:12
So we build a whole case.
1:16:13
We don't just rely on that.
Jason Savino
1:16:16
And if I could add to that as well, you know, it is so much more than just a match on a still in a sense that the investigators that are assigned are experts.
1:16:26
They look for physical characteristics first.
1:16:28
What am I talking about?
1:16:29
Scars, tattoos, anything that would match up the bridge to the nose, eyebrows.
1:16:35
They're looking at it from a physical standpoint, but then comes in the investigative aspect.
1:16:40
Right?
1:16:40
We need those corroborative factors to actually publish a match.
1:16:44
And what am I talking about with that?
1:16:46
I'm talking about, like Josh alluded to, is this individual out and available to commit the crime?
1:16:52
We'll look at the past arrests.
1:16:53
Have we seen it before?
1:16:55
Looking at body cam.
1:16:57
We're looking for the all those cooperating factors, maybe matches in clothing, matches to social media.
1:17:02
So it's so much more than just getting that still.
1:17:06
There's a physical and an investigative aspect to it as well, and the investigators are as good as they come at finding them.
Citymeetings.nyc pigeon logo

Is citymeetings.nyc useful to you?

I'm thrilled!

Please help me out by answering just one question.

What do you do?

Thank you!

Want to stay up to date? Sign up for the newsletter.