Diversity Training for the Police
I'm sitting on a bench in the Boston Common, about three miles from my apartment, looking up at The Embrace, a new, 20-foot tall sculpture inspired by a photo of Martin Luther King, Jr. hugging his wife Coretta Scott King.
The Embrace intends to honor "inclusion, racial equity, and justice", but it made national news for other reasons. Many people find it ugly and confusing. Some call it pornographic. In the words of comedian Leslie Jones, "you know it's messed up when Black people and the Proud Boys hate the same statue."
As I sit here, recalling the controversy, I find myself wondering: What would Dr. King have said? Not much, I suspect. One month ago, Tyre Nichols was murdered by five police officers in Memphis, the city where Dr. King himself had been slain. If Dr. King were still alive (he'd be 94 now) I imagine he would've spoken less about The Embrace than about Mr. Nichols, and about so many other victims like him – those we remember, and those we may have forgotten or never knew. Wikipedia, for instance, maintains a list, organized by year, of "unarmed African Americans killed by law enforcement officers in the United States". It is a very, very long list.
In this newsletter I want to tell you about a new study, published last week, examining the impact of diversity training on the attitudes and behavior more than three thousand police officers. To the extent that police violence against Black people and other minority groups is driven by negative biases, diversity training could, at least in theory, help reduce these biases and change behavior.
The scope of the problem
There are more than 800,000 sworn law enforcement officers in the U.S., and, as you'd expect with such a large group, they vary widely in attitudes and behavior. For instance, a 2017 Pew report found that even among White police officers, a substantial percentage acknowledge that police violence against Black citizens reflects a broader social problem rather than a handful of isolated incidents. Anecdotal evidence also hints that "The Police" is not some sort of monolithic racist entity. A good friend of mine in Dallas is a police officer, and one of the best doctoral students I've ever worked with was formerly a cop in Los Angeles. I haven't seen either one on the job, but they both strike me as exceptionally fair-minded and compassionate.
All the same, it's clear that law enforcement in the U.S. has a major problem with racism. Here's some of the evidence:
1. Studies have repeatedly shown that Black people are more likely than White people to be questioned, threatened, arrested, injured, or killed by police officers. These patterns, clearly documented by contemporary statistical data, have deep historical roots. In The Condemnation of Blackness, for instance, Dr. Kahlil Muhammad, traces racialized policing in the U.S. back to the antebellum "slave patrols" consisting of White citizens deputized to police the behavior of Black slaves.
The racist trope that Black people bring police violence on themselves is not supported by the evidence. For example, looking at a national database of incidents between 2013 and 2021 in which police killed a person, Dr. Reed DeAngelis at University of North Carolina found that White victims were much more likely to have been armed than Black or Hispanic victims (see figure below). The implications are chilling: Although Black Americans are more likely to be killed by the police than White Americans, Black Americans are less likely to be armed when they're killed.
Other studies point to implicit biases underlying police behavior. For instance, Joshua Correll at UC Boulder has shown that in video game simulations, police officers are quicker to shoot armed Black suspects than armed White suspects. However, police officers are also quicker to decide not to shoot unarmed suspects when the suspects are White. Correll attributes these findings to how readily (and thus how quickly) police are able to process stereotype-consistent information about armed Blacks and unarmed Whites.
Racism is not a purely individual-level phenomenon. Racial disparities in police violence vary widely from city to city, and systemic racism has been identified at the departmental level and beyond. A famous example comes from the Department of Justice investigation into the 2014 shooting of Michael Brown in Ferguson, Missouri. The DOJ's report received intense scrutiny for concluding that Darren Wilson, the officer who shot Mr. Brown, had not committed any civil rights violations. But the report also showed that systemic racism had long pervaded the Ferguson police department and broader community. In short, the DOJ concluded that racism was systemic in Ferguson, though not directly involved in Michael Brown's shooting. (Of course, it's hard to say exactly what had been running through Officer Wilson’s mind, and it's hard to prove the negative. A nuanced 2015 profile of Mr. Wilson does seem to hint at bias.)
In sum, saying that law enforcement has a problem with racism is like saying that St. Louis has a problem with violent crime. St. Louis certainly does have a violent crime problem (currently it ranks #1 among U.S. cities in that respect), but not every resident of St. Louis experiences such crime (the estimated rate is 2,082 incidents per 100,000 people), and the city also has some of the safest neighborhoods in the U.S. These aren't contradictory statements. The risk of being victimized in St. Louis depends on where a person spends time in the city – just like the risk of racially-motivated mistreatment by a police officer depends on where a person of color spends time and which officers they interact with.
In a way, President Biden summed it up fairly well in his State of the Union Address on Tuesday: "I know most cops are good, decent people...But what happened to Tyre in Memphis happens too often." Much too often.
One response to racialized policing: Diversity trainings
In recent years, spurred in part by reports of violence against minorities, most police departments have begun to require recruits and working officers to participate in diversity trainings intended to reduce bias. However, most of these programs aren't directly informed by research on biases and bias reduction, and few have been evaluated for effectiveness. These shortfalls motivated the study I'll be discussing.
The Psychological Science study
This study, published February 3 in the leading journal Psychological Science, was conducted by Dr. Calvin Lai at Washington University in St. Louis, and Dr. Jaclyn Lisnek at the University of Virginia. Lai and Lisnek surveyed 3,764 police officers before and after participating in a day-long, evidence-based diversity training. (Roughly three-quarters of the officers were male, and roughly three-quarters had at least five years of experience. The officers, and the people they reported working with on a daily basis, varied widely in racial/ethnic backgrounds.
The diversity training program
Managing Bias (MB), the diversity training program used in this study, was created by the Anti-Defamation League "to reduce the influence of bias in interactions and decision-making by law enforcement, contribute to improved police-community relations, and increase officer safety." The program is co-led by a pair of educators who focus on discussion and active learning exercises rather than lecturing. The goal is to help officers understand the nature of bias, where it comes from, and how it impacts behavior as well as relationships among police officers and community members.
The program also taught officers five evidence-based practices for reducing biased behavior: mindfulness (becoming more aware of the present moment), stereotype substitution (replacing negative stereotypes with positive mental images), perspective-taking (understanding other peoples' points of view), individuation (trying to know others as unique individuals), and diversity exposure (seeking opportunities to connect with people from different backgrounds).
Given the sensitive nature of the study, Lai and Lisnek did not reveal where it was conducted, though they reported access to a total of 251 training sessions conducted by 24 different educators. About half of the trainings were conducted face-to-face and accompanied by pen-and-paper surveys. The rest were administered online following the onset of the pandemic.
Survey Measures
Each officer completed three surveys, one at baseline (before training began), one at the end of the training, and one at followup (about a month later). Here are a few of the variables measured:
1. Understanding of bias.
Officers indicated the extent of their agreement with statements like "Whether I am aware of it or not, I use a person's race or ethnicity to form an impression of the kind of person they are".
2. Concerns about personal bias.
Officers indicated how much they agreed with each of six statements that began "I worry that I act in an unintentionally biased way toward...." followed by one of the following groups: Black people; Asian people; Hispanic or Latino people; transgender people; and gay, lesbian, or bisexual people.
3. Intention to use strategies.
The end-of training and followup surveys asked "How effective do you think these strategies are for addressing issues related to bias?", "Do you feel like you could execute these strategies correctly if you wanted to?", and "How motivated are you to use these strategies?"
4. Actual strategy use.
The five strategies taught during training were briefly described in the survey, and officers were asked about each one "Over the past seven days, how many times have you used this strategy?" (This question was included in the baseline survey too, so that Lai and Lisnek could measure how much the training affected officers' existing strategy use.)
Main findings
There was some good news and some bad news. More bad than good, unfortunately.
1. At baseline, understanding of bias appeared to be limited. Officers didn't strongly acknowledge the existence of biases that influence their behavior (the mean for these questions was 3.09 on a 7-point scale).
2. At baseline, the mean score for concerns about personal bias was 2.2 (on a 7-point scale) across the six minority groups – in short, officers didn't feel that they treat minorities in a biased way.
3. Understanding of bias increased significantly after training, and remained significantly higher on the one-month followup survey. This is the good news. Some learning seems to have occurred, although Lai and Lisnek didn't report the final means.
4. Concern about personal bias increased after training but slipped back to baseline levels after one month. That's terrible news (unless you assume that the officers were already quite "woke" and had no cause for concern, in which case it's merely bad news that training didn't increase their wokeness).
5. Intentions to use the strategies, as well as actual strategy use, were not significantly higher at followup than they had been on earlier surveys. Definitely bad news.
In sum, the day-long diversity training increased officers' general understanding of bias, but failed to have other enduring effects. As Lai and Lisnek put it,
"Our study converges with the mixed findings of diversity trainings found in other professional sectors such as companies or academia...These studies indicate that the current generation of diversity trainings are effective at changing minds but less consistent at changing behavior."
A closer look at the data
Studies on the impact of diversity trainings typically look at mean changes in key variables (e.g., strategy use). If there aren't significant mean changes, training is deemed ineffective. Although this makes sense, means can mask important differences among individual participants. In this particular study, it's possible that a subset of officers actually did benefit from training.
I reached out to Calvin Lai, the lead author of the study, to ask how many officers showed improvements in attitudes and/or behavior following training. In an email to me, he said that these analyses weren't run, but he added the following:
"[We did look] at individual characteristics of officers. We found big differences at baseline. Officers who were not White, not male, and younger tended to believe in the existence of racial bias and were more likely to be using evidence-based strategies to mitigate bias at baseline (e.g., perspective-taking). We also looked to see if any characteristics of officers were linked to greater learning immediately...To our surprise, we found that no characteristics were consistently learned to greater learning immediately. Officers were learning similarly regardless of their background."
The baseline differences Dr. Lai noted here illustrate my earlier point about variability in the attitudes and behaviors of police officers. The fact that officer learning was unrelated to their background tells us that the one piece of good news from this study may be relatively generalizable. Apparently no particular subset of officers learned the most from the training.
I still believe that data analysis should've, at minimum, identified how many officers improved, stayed the same, or got worse with respect to each variable measured. It makes a difference whether, say, 2% of officers began using more bias-reducing strategies versus, say, 14%. Both numbers are consistent with the lack of mean change overall. However, 2% would hint at random variation and imply that the training had no actual impact on strategy use, while 14% would hint that the training actually is effective, but, for whatever reason, only benefits about 1 in 7 officers. (Since training requires time and money, administrators need to know, among other things, how many trainees are likely to benefit. Perhaps 1 in 7 is enough to justify such trainings…?)
I don't fault Lai and Lisnek for skipping the kinds of analysis I'm calling for here. Over-reliance on means to evaluate interventions is commonplace in experimental research. I've written about this in other newsletters (see here, for example), and I'll probably devote a chapter to this topic in the book I'm planning.
Better diversity training
I also asked Dr. Lai what kinds of changes might've made the diversity training more effective. His answer was brilliant, and so, in spite of its length, I want to reproduce most of it here:
"At the end of the day, diversity training is just a form of teaching. So, many of the things that we know make teaching more effective will also make diversity trainings more effective. That includes:
Longer trainings, period. When we teach students math, history, or psychology, we often expect them to take a whole quarter or semester to learn it well. Why, should we expect anything different from diversity education in organizations? There are limits to what any single day of education can do, even when it is well-designed.
More practical skills. The training focused on teaching officers to see people as unique individuals. While that is generally useful, it may be hard to take that abstract idea and use it easily in daily work. Research shows that teaching skills that are closer to the behavior that you want to change is a way to guarantee effective behavior change. So, future training could make the skills more specific the daily work of officers (e.g, how someone might initiate a pedestrian or traffic stop).
Active learning with skills-based practice. Despite their best efforts, the training did not focus nearly enough time on drilling specific social skills and actions that officers could use on the job. We are currently testing a revised version of the training that will focus more on the applying the evidence-based skills they learned in scenarios based on real-life events.
Taking multiple trainings over a longer period of time rather than one big training. Memory research shows that learning that is spaced out over time is better retained than learning that happens in only one day. We actually have an upcoming project where we will test “booster shot” trainings so that officers can build the skills they learn into enduring habits.
Beyond that, there could be greater integration into policing organizations themselves. A common problem with continuing education or training is compartmentalization. You go in for one day out of the year, take the training, then never hear about the lessons of the training themselves. Sometimes you might even get counter-messaging from your peers or superiors. Any benefits of a training could wither away on that infertile ground…
I wanted to applaud after each one of those points. And, one more can be added. In the article itself, Lai and Lisnek made the following remark:
"We also found that sessions with the single ex-law-enforcement educator had better outcomes, although we hesitate to generalize too much from a single individual."
I appreciate their caution in not generalizing from a single case, but it wouldn't be surprising if it turned out that educators with prior careers in law enforcement and/or better pedagogical skills yielded better outcomes. The success of training depends on the characteristics of the trainers. In this particular study, it's unclear how well the training was delivered by the other 23 trainers. All we can say for sure is that training didn't help much on the whole.
Final words: The jolt of the new
Some of the tamest art you can think of was shocking when it first appeared – even Monet and Renoir were viewed as radicals and widely disparaged decades into their careers. As for The Embrace, my hope is that once people get over the the initial jolt, we'll come to see the sculpture as a symbol of vulnerability and security, and a source of inspiration.
With other kinds of jolts, it's best not to get over them. As a white person in my teens and 20s, I had a vague sense that Black people and other minorities were differentially victimized by police brutality. There were news reports. There were historical documents. There were the speeches of civil rights leaders like Reverend King ("We can never be satisfied as long as the Negro is the victim of the unspeakable horrors of police brutality.") There was Richard Pryor, one of the first to bring racism into mainstream comedy ("Police got a choke hold they use out here. They choke n****s to death...Did you know that? N****s going "yeah, we knew." Whites going "No, I had no idea."") All of this was informative but, I'm sorry to say, not very jolting to me as a young white person.
Then came the Rodney King beating in 1991. What's distinctive about that incident, and others since then, is that it was videotaped. The footage is horrifying and almost unwatchable, but even if you didn't watch, you surely heard about it. Nobody could fail to know. When George Floyd was choked to death, the White people Richard Pryor alluded to could no longer say "I had no idea." The challenge now is to avoid becoming numbed by the ongoing succession of tragedies, and to remain, as it were, jolted.
The statistical testimony is jolting too, if less visceral. American policing continues to be pervaded by racial violence. Given the extent of the problem, it seems unlikely that diversity training alone could be the solution. And yet, if such trainings were part of cultural changes within police departments, we might see progress. To that end, I close with one more comment from Dr. Lai in his email to me:
Imagine if police departments thought of training less as a one-off activity and more as an initiative that integrated with other organizational priorities! That could mean learning a set of evidence-based social skills or practices in a training that is then: (1) promoted at roll call and by superiors in your department; (2) practiced as part of daily work duties; (3) enshrined in departmental policies and regulations; (4) reinforced in hiring and promotion; and (5) highlighted as a core value of the department. Under those conditions, trainings wouldn’t have to do everything by themselves. They would be one part of a broader set of reforms within the department."
In short, change the culture, and the people may change too. One can hope.
Thanks for reading!