A 2014 “die-in” at George Washington University to protest police brutality. Photo by Jamelle Bouie.
Nearly every day in America, the list of victims of police shootings gets longer. Many of the people whose names are added to that list are Black, like Alfred Olango, a El Cajon, California, man who was killed last week. Many, like Olango, are also disabled—the facts of his case are still developing, but witnesses claim he had a seizure and may have had a mental health condition. But we don’t know exactly how many people are killed by police annually, nor do we know the specifics of their lives and deaths, despite efforts like The Guardian’s police shooting database, The Counted. This is a social issue. It is a social justice issue. It is unacceptable that hundreds of people in the United States are killed by police annually and we know almost nothing about them.
That’s because the information we glean is from news reports and voluntary disclosures from police departments. There is no protocol to collate data on shootings, let alone submit it to any kind of central database, ideally one maintained by the Department of Justice as part of its mandate to address both law enforcement and civil rights issues. To know that numerous people are being killed by police in obviously suspect circumstances is one thing—to not know the true scope of the problem is another. We also know there are easy ways to prevent this endless series of deaths, like clearly outlining use-of-force policies and providing training for police officers.
Since comprehensive government data isn’t available, the Use of Force project uses The Guardian’s The Counted database to determine what policies work to reduce shootings and deaths.
California is hoping to change the database problem with URSUS, an online utility law enforcement agencies across the state will have to use in order to comply with AB 71, which requires law enforcement to submit annual reports on fatality and bodily injury statistics to the state. URSUS is a law enforcement-only database, and the logistics of releasing reported shooting statistics to the public, whether through URSUS or another tool, haven’t been clarified yet. The open-source tool was developed by Bayes Impact, an organization that applies data science to social issues, in collaboration with the California Department of Justice. URSUS launched a demo version last week, allowing interested members of the public to take it for a little test drive as though they were law enforcement agencies preparing a report, which is exactly what I did.
What gets interesting about URSUS, of course, is the recorded perceptions at the time of contact and the demographics of the victim, which I explored by using the public demonstration to create a fake “incident.” I said my fictional victim was “perceived armed,” though I answered in the negative to the follow-up “confirmed armed” prompt. There are obvious flaws with collecting after-the-fact data. For police forces looking for reasons to justify shootings, it’s easy enough to claim that officers thought someone was armed, as was the case when, for example, Tamir Rice was shot to death within seconds of police arriving on the scene. When the question of whether a victim was armed becomes a cornerstone of a justifiable homicide claim, the difference between “perceived” and “confirmed” is important.
As for demographics, the form does provide room for multiracial individuals by allowing you to check several boxes, though the list of provided races is pretty limited, including just “American Indian, Asian Indian, Black, Hispanic, White, Asian/Pacific Islander,” and “Other.” A dropdown does list several options within “Asian/Pacific Islander,” like “Japanese” and “Samoan.” While this may cover a range of census categories and a broad sweep of racial data, it still leaves some significant holes. Given the huge factor that race plays in police shootings, we need harder data than that.
For gender, the form offers three radio buttons: Male, female, and transgender—yet another reminder that transness is considered an “other” or third gender. Unsurprisingly, nonbinary people are not represented at all; at a minimum, these options should be presented as a checklist, not a set of radio buttons, to allow for multiple selections, and a “nonbinary” option would cover a variety (though not all, because gender is complicated) of identities. This approach to gender is incredibly worrying for statistical analysis, as someone looking for, for example, data on the number of trans women shot by police in California will be unable to obtain that information readily through URSUS.
The report also asks about “signs of erratic behavior,” listing options like alcohol or drug impairment alongside evidence of physical disability, mental illness, or developmental disability. It seems a little odd to me to classify disabilities under “erratic behavior,” but in fact, “erratic behavior” is often why police receive disability-related calls. It’s also how disabled people are described in the media, as seen in the Alfred Olango case. What’s more troubling is that there’s no place on the form to confirm that a subject had a disability. That means, in turn, that if someone tries to compile data on disabled people who were killed by police, their search will come up empty—all we can point to in California is cases where police thought someone might be disabled (or presumably were alerted by bystanders, though there’s no specific field for this). Would Keith Scott, who died just a week before Olango, have fit under this criteria, or would his traumatic brain injury have been erased? Would Olango have fit under this criteria as well as “perceived armed”?
These are not merely academic questions. It’s hard to pull together statistics on the subject, though people have tried, but a very high percentage of people who experience police violence are disabled—perhaps as much as 50 percent, according to a report by David Perry and Lawrence Carter-Long for the Ruderman Family Foundation earlier this year. Many are mentally ill; the Treatment Advocacy Center found that people with untreated mental illness are 16 times more likely to be killed during a police encounter). Others are deaf or hard of hearing, like John T. Williams, a deaf Native American man killed by Seattle police in 2010. Others have cognitive, intellectual, or developmental disabilities, like Ethan Saylor, who was killed by police in Maryland over a movie ticket.
Like people of color, many disabled people have good reason to fear the police and have even warned friends and family against calling them in a crisis. This is a particularly large problem for mentally ill people, many of whom have been killed by law enforcement after they, or their families, called for help when they were in psychiatric distress. With so many slashes to mental health services, police departments have become first responders, and it is not going well. Some cities are starting to address this problem with mental health crisis teams that include specially trained officers and social workers, but it’s an upward battle.
It’s also a highly intersectional one, which makes URSUS reporting such a powerful potential tool. Many of the people of color killed by police are also disabled, and that is not a coincidence: Both people of color and disabled people are more likely to be shot by police, but disabled people of color sit at a dangerous intersection of social and law enforcement attitudes. Many people of color are also more likely to be disabled, a consequence of environmental racism, occupational disparities, and differential access to health care. As the shooting of Keith Scott, illustrates, it is critical to weave discussions of disability into the larger conversation about police shootings, but to do that, we must acknowledge that this issue exists in the first place, and that requires data. URSUS could be a radical force for highlighting the role of disability in police violence, but it must include a follow-up question: Was the victim confirmed to have a disability? If an officer didn’t think a victim was disabled, but she actually was, how do we account for that?
Am I just looking for things to be offended by? Yes and no. As a quality assurance engineer friend of mine likes to put it, “My job is to break things to make them work better.” URSUS has the potential to be an incredibly powerful tool if it is used as designed—which requires training departments and holding them accountable to ensure it actually is used. ‘Accountability is no small task, as seen with the Deaths in Custody Reporting Act, passed just two years ago, which advocates claim is already being manipulated to water down the data. It will be useful for the groups across the United States working to put a stop not just to police shootings, but to all excessive use-of-force incidents, which could reorient law enforcement to its intended purpose: to protect and serve the public.
This database is also, in a sense, a proving ground. If URSUS works, it can be used to pressure other states, and the feds, to enact similar mandatory reporting requirements and use databases built on the same engine. That means a rich wealth of data that could help us better understand the scope of the police violence epidemic in the United States and start picking it apart. But in order to do that, we need good data. Without it, we’re back to square one. It’s our obligation, and perhaps even our sacred duty, to break URSUS in as many ways as possible to highlight where it needs improvements.
Silicon Valley is familiar with app testing and agile development. The government is not. URSUS may prove to be a turning point not just in data collection, but in the very way the government conceptualizes technological tools.