Washington University’s Prison Education Project (PEP) hosted Marisa Omori, Ph.D., to speak about how the development of artificial intelligence, specifically facial recognition technologies, may impact racial inequality in the criminal justice system, April 11. 

Around 20 WashU community members gathered in Hurst Lounge to hear Omori speak in an event titled “Invisible Inequalities: Mechanisms of Racial Inequality in the Criminal Legal System,” including many faculty who work with PEP. The talk was a part of PEP’s annual Maggie Garb Lecture Series. 

PEP, established in 2014, is a WashU organization that strives to educate incarcerated people at the Missouri Eastern Correctional Center (MECC) and the Women’s Eastern Reception, Diagnostic and Correctional Center (WERDCC).

Omori is currently a Ph.D. Graduate Director and Associate Professor at the University of Missouri, St. Louis (UMSL). She specializes in racial inequality in criminal justice, courts, and sentencing, as well as in punishment and social control. 

She spoke about how facial recognition software, among other “formally race-neutral” institutions and platforms, often worsens racial inequalities in the system. 

“The facial recognition tools that have been used by law enforcement often perform worse on dark skin tones,” Omori said. “There is concern [these technologies] disproportionately matching mugshot photos, body camera photos, and other kinds of surveillance to databases in error.” 

Omori said that she collaborated with computer scientists to research the rates of misidentification of booking photos in different facial recognition software, comparing their effectiveness for both white and Black arrestees. They found that these software predicted Black people about 92% accurately on average and white people about 76.4% accurately. 

“These errors mirror a lot of the stereotypes that exist in the legal system, in the sense that the models are better able to identify blackness rather than whiteness,” Omori said.

According to Omori, technology like facial recognition is used in several stages in the criminal justice system, including identifying people who are suspected of a crime and during prosecution as evidence. Footage may come from police body cameras or even public cameras. 

Omori said that another way these technologies aggravate racial inequalities is through the placement of cameras. Many cities have higher concentrations of public cameras in historically Black residential neighborhoods, increasing the likelihood of a Black person being identified as a criminal.

“Technologies and norms [for their usage] are developing from the ground up,” Omori said. “For now, there’s very little or almost no oversight of these organizational practices because the law has not caught up with how this technology is being regulated.”

According to Omori, a recent Government Accountability Office report warned civilians about federal agencies’ usage of facial recognition technology. 

She said that another issue is that facial recognition software is primarily being developed privately, meaning that there is limited access to information about how results are generated or data about their accuracy. 

Omori added that some companies like Clearview AI have been accused of selling images to police departments without the consent of pictured individuals, breaching privacy laws.  

PEP Committee Chair, David Cunningham, a WashU sociology professor who attended the event, said that he enjoyed hearing an expert in criminology speak because it is not an area of focus at the University. 

“At WashU we have a sociology department that adopts a lot of these perspectives, but we don’t really focus on criminology as much as a place like UMSL does,” Cunningham said. “So to be able to bring someone here who can use the social science techniques that we see here and apply them directly to these kinds of questions and problems is really great.”

PEP Associate Director and Reentry Coordinator, Jami Ake, WashU professor in the Interdisciplinary Project in the Humanities, also attended the event. She said it was interesting to see the interplay between modern technological advancements and historical systemic issues.

“[It’s important] to see how technology, in particular, connects with some of the issues that a lot of people are becoming more familiar with in terms of mass incarceration and the ways that inequality works, and [how] that gets replicated in the tools that we start building,” she said. “Even people who are building [these tools with the intent of] making them neutral are not aware of all of the biases that are going into them.”

Omori delivered this lecture not just on the WashU campus, but also twice more to students in the PEP at MECC and WERDCC. She said that learning about the issues she discussed is an important step in combating the racial inequality that these devices may produce or aggravate. 

“I think we have an obligation to investigate and think about the ways that facial recognition design is not an accident of the system or anything like that,” Omori said. “It is baked into the design, and a lot of times, having this increased awareness can help us to think about how to push for additional regulations.”