Facial surveillance seminar
Bert Huang, Assistant Professor in the Department of Computer Science, moderated the first event in a series of seminars entitled, “Making real-world data science responsible data science” for computer science students on October 7. The series is directed by the National Science Foundation-funded TRIPODS Institute, a multi-department, interdisciplinary effort across Tufts University that focuses on data science.
“The main theme of the seminar series is that it will feature people who have crossed disciplines to adopt policy or real action to ensure the security of data science technology.” Huang wrote in an email to The Daily.
The seminar focused on facial recognition surveillance technology, with guest speakers Ben Ewen Campen, a current Somerville municipal councilor and researcher at Harvard Medical SchoolEarth Kade Crockford, the director of the Technology for Liberty program at Massachusetts ACLU. The two worked together to ban the use of facial recognition technology in police investigations in Somerville in 2019.
According to Ewen-Campen, the ordinance requires new surveillance technology to go through a public approval process before it is implemented.
“We passed what is called a Surveillance Surveillance Ordinance which says that any new surveillance technology that the [police department] … want to[s] to buy, [these requests] actually have to go to city council and have a public process and then get an affirmative vote from city council to buy or start using these technologies, ” Ewen-Campen noted.
Crockford explained how the process unfolded without much debate in Somerville, and has since spread to other communities across the state.
“I think [Ewen-Campen] in fact contacted me because our colleagues in San Francisco passed the first nationwide ban on the use of facial surveillance by the government. So [Ewen-Campen] was like, ‘Hey, I wanna do that to Somerville,’ [and] it was a relatively quick process, ”Crockford said. “We did it in six weeks or something, and then it really set off a chain reaction where other municipalities wanted to do the same, so we worked with people in six other municipalities across the state. to do it.”
After the ordinance passed in Somerville, the Massachusetts ACLU campaigns to vote on municipal bans and educate both the public and local lawmakers on facial recognition surveillance technology and the dangers posed without proper surveillance. Crockford explained in more detail the goal of protecting communities statewide.
“The blunt instrument of a total ban was what we thought was appropriate and necessary for law at the municipal level… to create the kind of regulation we believe to be appropriate,” Crockford noted.
Crockford explained the differences between three different forms of surveillance technology.
The first one, face surveillance, is defined by Crockford as “the application of video analysis algorithms using biometric identifiers to video data.“This can include reviewing video data to reconstruct someone’s movements over a period of time.
Crockford explained the danger of face surveillance.
“There’s no real way to ensure accountability and oversight, even though we had the best possible law that said, ‘the police can only activate [this] limited system circumstances, ”Crockford said.
The second technology is emotion analysis, which Crockford says is based on the idea that algorithms can determine how people feel based on their physical characteristics. Crockford stated that analyzing emotions is not a “legitimate science” because it is extremely difficult to interpret emotions based on someone’s facial expression.
“[Someone] maybe smiling, but they’re actually extremely nervous – people have that reaction. They may be telling the truth, but get agitated, because they are anxious because they are under police questioning ”, said Crockford. “The video analytics system may say, ‘Well, this person is dishonest,’ when in fact it isn’t – they’re just having a normal human response to a stressful situation. “
Crockford acknowledged that some legitimate applications of this technology existed, such as a system that automatically alerts truck drivers if they seem to be falling asleep.
The third form of facial recognition technology is known as image matching. This technology is used when an online post or video shows a crime being committed, after which a police officer can use a screenshot to get a clear image of the assailant’s face. They then take the image and other evidence to court and, for probable cause, seek permission to use facial recognition technology to identify the suspect.
Crockford believes that this technology is different from facial monitoring and emotion analysis.
“This poses less of a threat to our civil rights and basic civil liberties, because we are not talking about analyzing all the CCTV data,” [only] individual criminal investigations when a human being has identified a still image of a person… relevant to a criminal investigation ”, Crockford noted.
According to Crockford, Massachusetts has hundreds of different police services, which means that centralizing the image matching process can be effective in criminal investigations. Crockford proposed that the local police obtain a warrant to identify a suspect from an image, then take the image to the state police department, which has a facial recognition system.
“We think this is the right solution because when there is an algorithm that all police departments use, there may be real oversight and real accountability to make sure that algorithm is audited. independently. [and] has low rates of racial bias, ”Crockford said. “It is possible in this scenario that there is oversight of police use, as there is only one entity, the Massachusetts State Police, which is responsible for carrying out all this research… It is therefore the final regulatory solution that we are aiming for. . “