Federal Consent Decree Monitor Admits Facial Recognition Technology is racially biased.
Bias worst toward Black people and women if algorithm is coded by white men.
For some, hearing facial recognition has racial bias is like hearing water is wet. But, it has to be worth a bit more hearing it from one of the people appointed by the federal court to oversee the NOPD consent decree.
At a public meeting for the consent decree, the deputy federal consent decree monitor David Douglas said, “the technology itself is subject to all sorts of biases.”
Douglas admits facial recognition tech it very accurate as long as conditions are perfect, but he said that studies have shown “the conditions that lead to high accuracy are not often present in the real world.”
The deputy federal monitor also said the tech is “less accurate the darker your skin is, less accurate for women versus men and less accurate if the lighting is poor.”
If environmental and implicit technological biases weren’t enough Douglas says there’s one more form of bias more concerning than both.
The race of the programmer.
“The racial makeup of the people who make the algorithm that makes the (facial recognition) match, affects the bias in the matches,” said Douglas.
Douglas goes into an example saying that if the algorithm or “computer formula” is written by a white man it will be more bias against people of color and women. He quickly added that if the formulas are written by women or people of color biases show up, but in different ways.
However, that still leaves the “bias deck” stacked against Black people and women because the majority of people who work as programmers are white men.
The reason this came up at the recent public consent decree meeting, was at the time the 400+ paragraph consent decree was originally drafted, there were not yet discussions on the use of facial recognition technology.
Recently, by a vote of 4-3, the New Orleans City Council voted to remove the ban on the use of facial recognition technology for the New Orleans Police Department.
Douglas said the federal monitoring team was apart of the conversation every step of the way to instruct and approve how the tech could be used within the confines of the consent decree.
According to Douglas, facial recognition for the NOPD “can only be used to match a known actual photograph to help identify a given individual.”
For instance, under the current consent decree guidelines NOPD is not allowed to run a person’s photo through facial recognition software to track someone down in real time using live cameras.
According to the rules of the consent decree, NOPD “can’t use it to surveil a crowd of people just to see who’s there,” said Douglas.
This was tactic that now has the New York Police Department at the will of the New York Supreme Court. For what it worth, newly hired NOPD Chief Administrative Officer Fausto Pichardo abruptly quit in October following the protest after clashing with city officials who questioned NYPD’s excessive use of force during the protest.
Douglas said, given the scenario of a crime like carjacking, the NOPD is allowed to use a “known actual photograph” of someone who is accused of the crime. He said it’s a more modern technological way of going around showing a photo to a group of people to verify if the person in the photo is who police believe it to be.
Douglas also notes the evidence gathered by the technology wouldn’t be able to trigger any legal recourse, but instead be used as a lead to gather more substantial evidence.
The conversation around facial recognition technology made one time city council hopeful Bob Murrell ask how the local Independent Police Monitor is handling recent reports of NOPD getting assistance from a locally housed facial recognition company to “solve” a recent assault in the French Quarter.
According to the report, a local bar owner opted in to ProjectNOLA, a registered non-profit that claims to be “the most cost-efficient crime camera program in existence” to help NOPD ID a suspect.
In short, you’re able to give ProjectNOLA a photo and within seconds they are able to scan their cameras and find someone on their cameras.
Hearing of how all this went down led Murrell to file a report directly with IPM as well as question the federal monitors oh how they felt about this usecase by NOPD.
Douglas, who seemed to have learned of this from Murrell’s question says based on the facts of this case in particular it would be something worth looking into.
He said anything that even has the appearance of being an “end run around” the system is very concerning.
Anyone operating in good faith would agree the something has to be done about the crime in New Orleans.
If we’re together on that, we should still be together on agreeing facial recognition technology is, as folk like to say “no silver bullet” to solve crime issues in the City of New Orleans.
So, now tell me why it’s not best divest all the police funding for facial recognition technology from NOPD and put it directly into community programs with proven impact on growth and development.
Just a thought.