Data suggests ethnic minorities, particularly African Americans, Latinos, and Native Americans, are underrepresented in high-status careers such as STEM fields and the law. There are myriad reasons for this discrepancy, going back through the life span to educational disparities in the schools. Improving equality for all students is an important goal. Every student should be judged based on their skills and abilities, not the color of their skin or membership in identity groups. Universities have rightly implemented diversity trainings as one approach to open doors for underrepresented groups of students and foster unity among students of different backgrounds. But some trainings have arguably begun to employ jargon from the progressive left or employ concepts that have been challenged in the psychological literature or have unclear or negative effectiveness. It’s worth asking: What works and what doesn’t in diversity training?  

Significant controversy has emerged in the psychological literature regarding diversity training in educational settings and in the workplace. Put simply, many diversity training initiatives appear to fail. When done well, diversity training can promote greater cooperation among individuals of different backgrounds, foster understanding, and reduce unwanted or discriminatory behaviors. But, as with all things, research evidence sometimes changes, and some practices may become trendy despite evidence they are of limited value or even negative. On university campuses, advocacy groups and administrators may push ideologically loaded trainings that may ultimately contribute not to increased empathy but to a culture of fear, hostility, and intimidation.

Implicit Biases, Microaggressions, and Critical Race Theory

A significant industry has sprung up offering trainings on “implicit biases” that have found their way into both educational settings and other workplaces. Even the Department of Justice announced such a program in 2016 that is “…based on the latest social science research.” This concept refers to biases that, unlike explicit prejudice, may be unconscious yet nonetheless affect relationships among different groups. A highly publicized example was Starbucks’ decision to implement a day of implicit bias training following a racial incident at one of their stores. Many outlets at the time wondered if Starbucks’ training would work. Most such queries regarded whether a single day of implicit bias training would be enough, which is fair. A better question would be whether such training helps at all. In recent research, the utility of the concept of implicit biases has come under question. Put simply, recent evidence suggests that implicit biases have little impact on real-life behaviors. If such trainings are mandatory, they might even be harmful.

One of the larger efforts around implicit bias training has involved the New York Police Department’s investment of $4.5 million to offer implicit bias training to police officers in hopes this might reduce racist incidents among them. An empirical evaluation of the training suggested that officers reported learning content, but that it ultimately had little impact on actual officer behaviors. Thus, implicit bias trainings may be popular (and lucrative), but there is little clear evidence yet that they work.

Another controversial issue revolves around trainings focused on microaggressions, which are particularly popular on university campuses and often part of first-year programs. Microaggressions are slights, often unintended, that highlight the differentness of people not in the majority group. As with implicit bias, the concept of microaggressions has gotten considerable attention in the public. However, recent scholarship reveals that the concept is not built on a solid foundation of scientific evidence and trainings, which suggests it may actually do more damage than good in polarizing different groups. In essence, people may become more primed to interpret ambiguous events as aggressive, become more upset by these events, and be less forgiving of minor or accidental transgressions after being informed to scan for and interpret them negatively.

Perhaps few areas of diversity training have been more controversial than those based, at least loosely, on critical race theory (CRT). For this essay, I include under this umbrella efforts influenced by the antiracism of Ibram X. Kendi, as well as the racialized/identarian approaches based on concepts such as “white fragility,” which even proponents tacitly acknowledge often degenerate into acrimony. Such approaches have been criticized for being reductive, misinformative, and incurious to data, and themselves bordering on racism toward both whites and people of color (particularly people of color who disagree with CRT). Such trainings may ironically resemble some of the worst aspects of past racist structures, such as separating/segregating people by racial or other identity categories, casting people as “good” or “evil” based on identity characteristics, implying that certain traits such as hard work and rationality are inherently “white,” advocating racial preferences in the deliverance of health care, or identifying any disagreement with CRT as “white fragility” or “multiracial whiteness” (for people of color who disagree with CRT) as itself evidence of racism. This is an approach I call “Fingercuffs Theory,” when any criticism of the theory is itself taken as evidence for support of the theory. 

Such practices are almost certainly harmful. The recent blowup at Smith College provides one example of how they can worsen race relations and send a university into turmoil. At Smith, as is often the case, the burden of harm fell on low-income workers with few protections who were falsely cast as racists. The culture of fear I spoke of earlier occurs when those who speak out regarding their concerns are then subject to “cancellation” campaigns, which occurred for a professor at the University of Chicago (though thankfully the university refused to give in to mob demands). I suspect some of these trainings will end up in the courts on both racism issues and hostile work environment issues.  

We Need Better Data

The simple reality is that we need better research on what works in regard to diversity trainings. Diversity trainings involve a massive industry that is often flying blind, too often pumping effort into trainings that are trendy among progressives rather than those demonstrated to be effective. Under the current cultural milieu, many universities are likely to look for DEI trainings and are just as likely to fall victim to bad trainings as other businesses. This is also a morally valenced and emotional field that is going to exacerbate the already high propensity for false positive results in social science. 

Ultimately, the data we have on what works in diversity trainings is not sufficient. Particularly as there is so much emotion, ideology, and advocacy built around this topic (understandably so), the data must be protected from the wishful thinking of those with axes to grind. This is a common problem for social science, hence the replication crisis, which has rocked our fields over the past decades. We have some data on IAT trainings, but much less on CRT-based trainings or “white fragility” trainings, or similar approaches that lean heavily into progressive academic jargon. Ideal studies would be randomized controlled designs. Such studies should be preregistered, and conducted by competent and trained clinicians, using standardized assessment instruments. As much as possible, participants should be shielded from researcher promptings to respond to assessment instruments in ways that are artificially consistent with hypotheses. Such programs could examine the effectiveness of CRT-based programs with those that, instead, focus on shared humanity, shared empathy, equal footing, and empathy across all identity groups. By preregistering, scholars should make a commitment to publish their results, even null results, whatever the outcome. 

The Failure of Professional ‘Science’ Guilds

Our main psychological organizations, such as the American Psychological Association and the British Psychological Society, have remained largely silent on these issues. Or, perhaps more accurately, they’ve taken a reflexively “woke” and, I’d argue, antiscience position of not questioning the dogma of CRT at all. Although the APA referred to evidence-based trainings in its statement on Trump’s executive orders to restrict certain topics for federal diversity trainings, it failed to mention that Trump’s ban (counterproductive though it was) specifically targeted those trainings with dubious empirical validity. This puts organizations such as the APA and the BPS in the unsavory role of at least implicitly promoting diversity trainings of either no known validity, or for which there may be reasonable expectations they may cause more harm than good. This seems an odd stance for organizations that are ostensibly science-oriented organizations.

That such organizations are failing is, frankly, not surprising. First, these organizations are professional guilds, not actually science organizations. It has long been known that they tend to promote views that benefit the profession or that simply conform to the worldviews of their dues-paying members. Much of this long-standing criticism has focused on the APA, for which deeply misleading policy positions have become commonplace, and which was even accused of altering its ethics code to allow psychologists to participate in harsh military interrogations (i.e., torture). The BPS has generally avoided this level of controversy, although a recent edition of its Observer magazine created uproar when comments by a once skeptical psychologist appeared to get “canceled” after a Twitter-driven outrage wave (note: I feel empathy for the editor who, I suspect, was in an unwinnable situation). The BPS, like the APA, seems to have increasingly eschewed careful, nuanced science for uncritical woke advocacy. On claims regarding its own institutional racism, the BPS chief executive claimed, “Are we institutionally racist? I think my answer would be that, if it feels like we are, then we probably are,” which is as dataless and unfalsifiable a statement as one can get. 

Ideally, our science organizations should be as honest as possible about complex, nuanced data, which often defies either simple explanations or close adherence to advocacy causes. By failing to do so, such organizations are increasingly at risk for contributing to real harm, including to the very communities they ostensibly are attempting to advocate for. It can be hard to push back against an availability cascade, but in failing to do so, they may actively make a bad situation worse, even if it is less threatening for those organizations in the short term. This means that organizations like the APA and the BPS should be honest when some approaches either are untested or (as in the case of IAT trainings) simply don’t work. Declining to do so, despite convincing evidence, often published in their own journals, is dalliance with antiscience, not science.

So, what kinds of trainings are most effective? First, when discussing controversial topics (such as microaggressions), it is best to acknowledge where uncertainties exist rather than push a particular ideological view. As a general rule, the more new words or concepts attendees have to learn, the more likely an ideology is being pushed. Second, trainings that allow for mutual respect and sharing rather than presenting some groups as “good” or “bad” (as with the “white fragility” concept) are likely to see more progress. Evidence suggests that trainings that promote empathy toward others reduce hostile behaviors. Trainings that are voluntary (possibly with positive incentives) rather than mandatory, focus on skills development, involve perspective taking, and are part of a long-term plan tend to be most effective. Put simply, effective diversity training generally focuses on the positive, bringing people together and informing them how they can be part of long-term goals that will benefit everyone. Trainings that focus on politicized ideology and polarizing jargon, what people are doing wrong, or trendy theories are less likely to succeed and may actually do harm.  

I wish to be clear that evidence-based diversity trainings can be helpful, and I fully support them. Programs that increase intergroup contact (as opposed to segregating by race, which has inexplicably become trendy in some circles) and foster cooperation on projects show some promise. Yet, more data is clearly needed. We need to better understand what approaches work to bring people together across differences and unify them. This means we need more preregistered, transparent, high-quality science than what we have now. In the meantime, pushing trainings for which the effectiveness is poorly understood (or for which even advocates acknowledge create significant hostility and acrimony) may cause real harm, including to the vulnerable communities we are ostensibly seeking to help.