A bill banning the use of facial recognition software will soon appear before City Council as Madison’s Public Safety Review Committee recommended the adoption of the ordinance Wednesday night.
The ordinance, which passed 4-1 with one abstention at the committee, would heavily limit future use of the surveillance software by government bodies. Concerns were raised with how the legislation would impact the Madison Police Department, which utilizes the software in investigations.
The vote was preceded by a public forum, in which over 100 Madison residents participated. Their opinions during this forum were varied, ranging from wholehearted support to vehement displeasure at the proposed legislation.
Three main themes emerged during the forum: Concerns regarding government overreach, and specifically police abuse of facial recognition technology in the future; potential false positives as a result of racial biases in the algorithm used by the software; and the fact that, if passed, this bill might hinder human trafficking investigations currently underway in the city.
Resident Kim Richman said that the proposed legislation was personally motivated and that the police should be allowed to fight crime as they see best.
“This should be used for public safety, not personal agendas,” she said in reference to the council’s vote. “Don’t accuse these officers of jumping to conclusions, which you all do on a regular basis.”
Gisela Wilson, a University of Wisconsin-Madison researcher in the Department of Integrative Biology, cited her expertise on the issue as a driving force behind her support for the bill.
“I have read 30 to 50 books on this issue,” she stated. “Facial recognition software databases, which are built without people’s consent, [are] in violation of the Fourth Amendment.”
Wilson also called into question the ability of police too, even outside of these abuses, identify the racial biases baked into this software.
“Police and officials have little understanding of biases and assumptions of the databases and the machine learning algorithms used to make these identifications. That means incorrect and biased info could affect people's lives, and contact with the police can be extremely dangerous,” she concluded.
Jack Phillips, an engineer at UW-Madison, pleaded with the council to accept the resolution, citing grim possibilities for future uses of the technology.
“Imagine a child growing up in a grim Orwellian surveillance state, being wrongfully identified by an algorithm trained on white tech workers because it can’t differentiate between people of color,” he said.
Interim Madison Police Chief Vic Wahl and Detective Julie Johnson, the head of MPD’s Special Victims Unit, appeared at the committee meeting to voice their opposition to the proposed legislation. The two law enforcement officials said repeatedly that, if passed, this bill would hinder their ability to effectively continue their human trafficking investigations.
In the presentations given by Wahl and Johnson, the subject of accuracy was repeatedly touched on. Wahl stressed that although some recognition software had high rates of inaccuracy, especially among Black and Brown individuals, not all did, citing the multitude of available programs. He also referenced a 2019 study by the National Institute of Standards and Technology, which noted “massive gains in accuracy” in facial recognition software since 2013.
Johnson, speaking directly after Wahl, explained how and what the Special Victims Unit does and its collaborations with the National Center for Missing and Endangered Children and Thorn, the main associates of MPD that use facial recognition software. When the Madison Police Department obtains a photograph of a missing individual, Johnson explained, they send it to these organizations in the hope that they will be able to produce leads by matching the photo with a one online.
Both parties stressed the importance of facial recognition in these pursuits, repeatedly calling Madison a hotbed for human trafficking as Dane County experiences a surge in the number of cases.
According to sponsors Ald. Max Prestigiacomo and Ald. Rebecca Kemble, the bill was written with the intent that MPD’s work with facial recognition software to aid in human trafficking investigations would not be hindered.
Kemble, in discussing her reasoning behind sponsoring the bill, described a 2-year long investigation with other alders about the potential impact of surveillance and facial recognition technologies on the community. They learned that, though recommended, there were no organizing regulations or organizational requirements for these services. Kemble said that facial recognition first came to her attention around 2018, during a committee on another surveillance issue. They decided at this juncture that facial recognition warranted its own committee.
Kemble also stated that the point of this legislation was preventative.