Certain cities and communities are watched much more closely than others.
Robert Williams was the first American to be arrested for a crime he did not commit based on the judgement of a facial recognition software. In January 2020, police officers drove up behind Robert as he was parking at his house and blocked in his car. They handcuffed him in front of his wife and two daughters. When they took him to the police station, they showed him two blurry photos of a heavyset man, dressed in black, and wearing a red St. Louis Cardinals cap, standing in front of a watch display. Five watches worth $3,800 were shoplifted. When the detectives asked if that was him in the picture, Robert responded, “No, this is not me. You think all Black men look alike?”
Surveillance and the use of facial recognition tools have grown exponentially in recent years. While The New York Times and other technology & law experts believe that Robert’s case is the first of its kind, it will certainly not be the last. As such, I developed a Python web-scraping tool to identify every single relationship that a police department has with Amazon Ring.
I found that these surveillance systems are disproportionately in Black neighborhoods by a factor of 12. In fact, the 5 counties in America with the highest Black populations have 120 police departments working with Amazon Ring whereas the 5 counties with the highest White populations have 1/3 that amount.
In 2018, Amazon bought a doorbell company called Ring for $1.2 billion. Ring smart doorbells are equipped with video camera surveillance. When Ring cameras detect motion, they start recording whatever is in view whether it’s a burglar, passerby, or Amazon delivery person. This video is sent to an app on the homeowner’s phone, which can then be shared publicly or privately with direct messages. Amazon has another business line called Rekognition which sells facial recognition technology.
1,752 police departments across the country have created formal partnerships with Amazon Ring. The first Ring-Police partnership was announced on March 22, 2018 with the Greenfield, Wisconsin Police Department. Since then, police have made 5,932 requests for video surveillance information from Amazon to access Ring doorbell footage.
Surveillance exists in many different forms. State and local governments may maintain security cameras, private companies may have their own CCTV feeds, and many apps track our location. However, the Ring-Police partnerships are a completely uncharted space for surveillance based on the scope and public-private nature of the engagement.
Black communities surveilled more
Nearly 60 years ago, the FBI authorized wiretaps for Martin Luther King Jr.’s home and the Southern Christian Leadership Conference (SCLC) offices. 60 years before that, Plessy v. Ferguson was decided in the Supreme Court to keep Black people in their separate communities and under close supervision from local law enforcement. Today, Black communities are still surveilled at much higher rates than White Communities, leading to increased tensions and criminalization.
A 2019 Motherboard investigation revealed that Ring exacerbates existing racial injustice within neighborhoods. People of color are disproportionately reported as “suspicious” even in the neighborhoods they live in and have their photos uploaded for public comment. Users go onto the app that comes with Ring called “Neighbors” to complain about Black and Brown delivery drivers not doing their job well enough and commenters often respond to reports of petty crime (75% of reported crimes are package theft) by encouraging calling the police (ignoring the many high-profile cases police brutality) or making racist and threatening comments themselves.
In fact, I found that Police-Ring partnerships are disproportionately in Black counties. The average US county has a Black population of 13,000, but for counties with a police-Ring partnership the average Black population numbers 157,000. The average US county is 9.1% Black, but for counties with a police-Ring partnership the average is 12.3% Black.
Cook County, Illinois has the most police-Ring partnerships of any county in America and is also the county with the most Black people. The county includes Chicago and its Southside and is home to 1.2 million Black Americans. According to Amazon, 71 police departments have created partnerships with Ring in Cook County.
In 2014, Chicago police officer Jason Van Dyke killed Laquan McDonald, a 17-year-old Black man. In 2018, Van Dyke was found guilty of second-degree murder and 16 counts of aggravated battery. Later in 2018, the Chicago Police Department added 30,000 cameras to local streets. The following year in 2019, the first police department in Cook County created a police-Ring partnership. Since then, Cook County has added 3 new police-Ring partnerships every single month.
There is little evidence that Ring reduces crime. Researchers from MIT found that there is little evidence to support Ring’s claim that it reduces crime and the company does not publish data on the subject. Ring denies facial recognition is in the works, yet leaves open the possibility of implementing facial recognition and refuse to commit to never selling biometric information.
Facial Recognition and Surveillance
More than half of all Americans are now in facial recognition databases.
Several cities across America have banned facial recognition due to its untested governance and controls. San Francisco, Boston, and Portland are among dozens of cities that have banned this technology, citing risks of “biases against Black people, women, and older people.” Research from Joy Buolamwini at the MIT Media Lab has shown that facial recognition softwares have error rates of 34% in darker skinned women, often because the underlying databases that train the algorithms are 77% male and 83% white. The National Institute of Standards Technology (NIST) analyzed 189 facial recognition algorithm from 99 developers using 18.3M images of 8.5M people and found that the algorithms were 10x to 100x more likely to misidentify African Americans, Native Americans, and Asian Americans compared to White faces.
Why has law enforcement pursued more surveillance?
Law enforcement officials argue that there are two pros to using facial recognition technology — improved efficiency and improved optionality.
- Improved efficiency — The Department of Homeland Security has argued that facial recognition technology can help the government more quickly screen passengers at airports and process people at immigration checkpoints. This process is currently very manual, and often requires a federal agent holding up a picture of your face (either a passport or driver’s license) next to your actual face and looking back-and-forth to see if they are similar. In a small footnote in its November 2020 report, the Customs and Border Patrol has said that for Global Entry terminals, “Facial recognition transactions have reduced kiosk processing time, already short, by an astonishing 90 percent.”
- Improved optionality — In 2019, New York’s police commissioner James O’Neill wrote an Op-Ed in the New York Times arguing that “facial recognition makes you safer.” After explaining that he previously used ink-finger print cards and Polaroid cameras for mugshots, he goes on to say that facial recognition is just the next step in a long-line of evolution of police tools. “Technology has improved the profession beyond what the most imaginative officer could have conceived in those days. These innovations include facial recognition software, which has proved its worth as a crime-fighting resource since we adopted it in 2011… We use these methods solely to fill in missing or distorted data.”
The costs associated with paying for Amazon’s facial recognition software are remarkably low — the Washington County Sheriff’s Office in Hillsboro Oregon pays just $7 per month for the service. Nevertheless, the 2020 police budget included $600,000 for “smart policing technologies”.
The Path Forward
Courts need to rule on evidentiary standards — Senator Ed Markey has noted that “Ring has no evidentiary standard for law enforcement to request Ring footage from users” and Ring appears to use standards that have no basis in legal standing. For example, Ring currently limits requests to “video recordings only from an area between 0.025 and 0.5 square miles (0.5 square miles is approximately 10 city blocks)” from an event. This seemingly random distance may be wholly inadequate in rural regions or may be too expansive in urban regions, putting people at risk of the “plan view” doctrine. In the meantime, Ring can improve its Informed Consent notices so that when a police-request does come to a Ring user, they know that they don’t have to turn over the footage.
Create better transparency reports — Law Enforcement Ring Request Reports will improve transparency into police-Ring partnerships, thereby supporting greater accountability. Although Ring currently publishes a map on how many data requests it receives from police departments, this information is not detailed enough and difficult to navigate. Instead, Ring and police departments ought to follow transparency best practices for law enforcement data requests as illustrated by the Bureau of Justice Statistics. BJS has outlined 6 categories of statistics that law-enforcement should publish to show how effective their policies are (and whether they actually reduce crime). President Obama’s Task Force on 21st Century Policing (2015) recommended that police leaders work to “establish a culture of transparency and accountability in order to build public trust and accountability.”
The day that Robert Williams was arrested for faulty surveillance and facial recognition, his wife called his employer to say he had a family emergency — it broke his four-year record of perfect attendance. The following day was his 42nd birthday. As Robert sat in the interrogation room, the cops leaned back in their chairs, looked at each, and said, “I guess the computer got it wrong.” 30 hours later, Robert was released on a $1,000 personal bond. He stood outside the Detroit Detention Center in the rain for 30 minutes before his wife could pick him up. His daughter was waiting up for him when he arrived back home at 10pm and ran over to give him a big hug. Robert says that his daughter’s favorite game now is cops and robbers. She runs around the house, accusing her father of stealing things, threatening to lock him up.