Thirty police officers sat contemplating a deceptively simple question: At the scene of a car crash, why is it that the man dressed in a suit and driving a shiny BMW is usually more believable than the man wearing muddy jeans driving a pickup truck? It was late in 2018 at the New York Police Department, and under mandatory orders, they were there to learn about something lurking within everyone’s subconscious—racism.

While you may not show explicit signs of racism against any particular group, or even consciously think negative things about them, you probably still carry prejudice in the form of implicit bias. Implicit biases are a symptom of grouping, which is something our brains start doing shortly after we’re born. When someone says peanut butter, jelly comes to mind. Salt and pepper. Beaches and sand. Every day, we bathe ourselves in news reports, encounters with friends and co-workers, movies, books, and gossip. Each of these minuscule encounters accumulate and provide our brains with data about which people and things and adjectives tend to come bundled together, solidifying associations in our minds. That’s how we form implicit biases—whether we want them or not.

“We’re swimming around in this sea called culture, and we are creatures of that sea,” says Mahzarin Banaji, a professor of psychology at Harvard University. “Our bodies have become adapted to the salt water, and now this stuff is in us. True, I didn’t put that stuff there, hey, I don’t even want it there. But it is there.”

These associations are crucial in helping us fathom the world. In early childhood, babies begin crafting mental links between objects and their uses—for example, spoons are used for eating mashed veggies and fruit, and galoshes go with rainy walks. The connections become more nuanced and varied as we grow up to help our brains make sense of objects and scenarios we encounter. But as essential as these associations are, they also have serious consequences for our perception of other people, which in turn affects how we treat them. That can have disastrous effects on policing.

“The most common [example] is the black-hyphen-crime association,” says Lorie Fridell, a professor of criminology at the University of Florida. Many Americans grow up watching a disproportionate number of Black people getting arrested on television. According to a report published by Media Matters in 2015, between August 18 and December 31 of 2015, the evening news programs for WCBS, WNBC, WABC, and WNYW identified Black people as suspects in 68 percent of reports on murder, 80 percent on thefts, and 72 percent on assaults, although the actual statistics throughout 2015 were significantly lower, at 62.8, 69.2, and 52.4 percent, respectively.

Research suggests that most people carry an unconscious sense that Black people are more closely linked to crime than those of other races, which is likely at least partly to blame for the outsize danger Black individuals face when encountering law enforcement.

How implicit bias training works

Police brutality in the US is on the rise, with 1,004 people dying at the hands of police in 2019, 23.4 percent of whom identified as Black. To combat this, police departments in New York, Austin, Los Angeles, Miami, Milwaukee, and Minneapolis have contracted with a company called Fair and Impartial Policing over the last few years, spending millions of dollars for courses that claim to help cops shake off unconscious prejudice. In 2018, the NYPD spent $4.5 million to train all 36,000 New York-based officers and commanders for around two years, using a mandatory eight-hour module that aims to help the police understand their biases based on conversations about their own experiences in the field, and implement unbiased, controlled responses to emergencies. But experts say these interventions are often poorly received, and applying the science of implicit bias is far from straightforward.

“When they walk into the classroom, the officers are somewhere between defensive and downright hostile,” says Fridell. “They think we’re gonna shake our fingers at them and call them racist.” The first step in the training process is to get over that “defensiveness hump,” she says, and guide the officers to confront the ways their biases may have controlled their previous actions. Beyond that, a series of lectures and group discussions led by the instructor—always a former member of the force—aim to teach attendees that conquering implicit bias can lead to fair and effective policing of communities. Classes for sergeants and commanders focus on fostering a positive culture within their departments, weeding out explicitly racist individuals during the recruitment process, and helping subordinates avoid situations where implicit bias may get the better of them, such as pulling over ‘suspicious’ vehicles to search for drugs.

Fridell says that officers’ attitudes, knowledge, and skills tend to shift after training, at least in the short term. But detecting a course’s impact is difficult. “It’s hard to track what cases of excessive force or complaint didn’t happen because we had this training,” says Fridell. So how do we know if it works at all?

A heated debate

Mahzarin Banaji stumbled across the idea of implicit bias as a grad student in psychology at Yale University in 1990. Along with Tony Greenwald and Brian Nosek, she published a series of implicit attitude tests on Yale’s website in 1998. The tests measure the strength of associations between different concepts (black faces, white faces) and evaluations (good, bad) by asking test-takers to categorize words or images by pressing certain computer keys. The time taken to respond to various combinations reveals these users’ subconscious mental associations.

Banaji predicted they’d have data from 500 test results in one year. But in barely one month, they were overwhelmed with 45,000 submissions. “The Yale website would crash regularly in the first few days,” she says.

Over time, the team of three published other tests, which measured biases surrounding age, disability, religion, gender, weight, and sexuality, among many others, creating not only a 20-year data pool, but a worldwide phenomenon that went viral, and stayed viral. The data shows that an overwhelming majority of test-takers exhibit implicit bias against Black people, but the trend has been slowly decreasing, especially from 2012 onwards, during the years of the Black Lives Matter protests.

Since 1998, implicit bias has become a household name, given a spotlight in hundreds of research papers and dissertations using Banaji’s enormous pool of IAT data, an episode of Oprah, and former President Barack Obama’s Task Force on 21st Century Policing Implementation Guide. The guide, published in 2015, recommends the undertaking of “trainings and organizational change that address procedural justice, implicit bias, and de-escalation/use of force” and officers’ use of body cams. The popularization of the concept has spawned numerous implicit bias police training sessions, which have led to dubious results.

Take Minneapolis. City police have reportedly used force—kicking, punching, neck holds, and tasers—11,500 times since 2015. In 60 percent of these incidents, the subject was Black. Former police chief Janeé Harteau, who served the city from 2012 through 2017, brought Fair and Impartial Policing to the MPD in 2014. “I think it’s important that people understand that police officers are humans, they have unconscious bias,” says Harteau. “Not because they’re police officers. We hire from society. If there’s a percentage of racism, there’s gonna be that percentage in the organization.” Harteau asserts that this kind of training is essential. Having grown up in Duluth, Minnesota, Harteau had very little exposure to the Black community as a child. When she moved to Minneapolis to become a police officer, she entered a completely new world of high crime and poverty rates. “Had I not had the character and mindset I did, I might have had a different way of thinking,” says Harteau. “Even today, we all have to check ourselves. We all have a bias of some sort.”

The department’s 1,000 officers spent six or seven months undergoing training, resulting in “good feedback from rank and file,” according to Harteau. However, in the years following 2014, Minnesota only saw an increase in police brutality. On May 25, 2020, Officer Derek Chauvin killed George Floyd on Chicago Avenue by kneeling on Floyd’s neck for nearly nine minutes—the agonizing spark that gave rise to ongoing anti-police brutality protests across all 50 states.

“People are saying, ‘if Officer Chauvin had implicit bias training, why did this happen?’” says Harteau. “It’s hard to prove or track the crimes you prevented.” But she maintains that the training is necessary.

Many people think otherwise. Calvin Lai, assistant professor of psychological & brain sciences at Washington University, views the results of implicit bias training as inconclusive. “We simply do not know. There has been no published research on implicit bias training in police departments.” The results can vary, according to Lai—it can help certain individuals, but it can also completely backfire, especially in situations when officers are forced to attend, which tends to make them defensive. Dave Bicking, a volunteer and leader at Communities United Against Police Brutality in Minneapolis, looks upon this training with distrust. “Implicit bias was one of a whole package of reforms that was meant to look good without changing anything fundamental,” says Bicking.

Bicking, who has spent years fighting for dramatic police department reform in Minneapolis, sees explicit racism as more of a modern problem than cities might like to admit. “You can see it, hear it in their language, see it in statistics in how many Blacks versus whites are victims,” says Bicking. Implicit bias training won’t improve the behavior of cops who are consciously aware of—perhaps even perfectly comfortable with—their racial prejudices. And mandated attendance could feed a sense of conflict they feel against the communities they patrol. Another one of Bicking’s reforms targets the police training system, and how it exacerbates the “us vs. them mentality.” In one 2018 study, researchers claimed they saw this mindset grow more acute with each stage of police training.

Long Island-based civil rights attorney Frederick Brewington points out that police training is not just inadequate—it’s also alarmingly brief. “They get six months and they’re handed a gun,” says Brewington. “If we’re putting issues of life and death into the hands of certain individuals, why do we only give them six months of training in an academy?” In Minneapolis, the police academy takes between three and four months. The lack of consistent psychological profiling of officers is another dangerous issue, according to Brewington. Officers see “some of the worst things that humans can do”’ but only receive one mental evaluation before they enter the police force. “The way we do policing in America, we’re at a point where we can’t do it patchwork,” he says. “We’ve gotta disassemble the process and recalibrate.”

How can we improve implicit bias training?

Implicit bias training may not be an easy panacea, Banaji says, but it shouldn’t be scrapped entirely. There are several key opportunities for improvement—especially when it comes to making the process more palatable to participating officers.

First, the training should not be mandatory. Eugene Borgida, a professor of psychology and law at the University of Minnesota, claims that forced participation tends to make participants defensive and untrusting, which makes for a less effective program. Slowly gaining the respect and interest of a few officers at a time may be a better strategy than mandating the process for an entire department.

In 2019, Banaji personally oversaw the voluntary implicit bias training of officers in Cambridge, Massachusetts, by inviting more than 300 to her class. Just a third of them chose to show up. Banaji started the session with an open, vulnerable admittance—that she showed signs of racial bias on her own implicit attitude test. “This is the least I can do, to show them that we are speaking about something that is true of all humans,” Banaji says. “I gave them hands-on experiences and data. I gave them optimism at the end about change, and let them go.”

Two weeks later, Banaji picked up a phone call from the Deputy Superintendent at the Cambridge Police Department. Based on positive feedback from those who had attended the first session, the other 200 officers now wanted to participate. After five meetings, Banaji surveyed the attendees on the helpfulness of the course. She found that, on a scale from 1 to 5, scores increased from 1 to 4.5 throughout their time together. Banaji stands by these trainings, claiming that methodology rooted in her research on bias can help improve efficacy. “It can be done right, if you root it in the science,” she says. She adds that research on the subject is more essential than ever, especially as the Minneapolis city council has announced its plans to defund and dismantle the city’s police department. The restructuring of these systems should be informed by careful reasoning and research.

Banaji regularly thinks back to a lecture Martin Luther King, Jr gave to the American Psychological Association just weeks before his assassination in 1967. “He said, ‘you aren’t making policy. But what you can do to help the movement is by telling it like it is,’” says Banaji. “We cannot do everything. We cannot make policy or organize communities or dismantle any department. But what we can do is reveal who we are, how distant our behavior is from our own values, and then hope that this disparity will shape who people will choose to become.”

Correction: A previous version of this article mistakenly stated that 195 civilians were killed by police in Minneapolis from 2000 to 2020. In fact, 195 civilians were killed by police in all of Minnesota during this time. We regret the error.