Prejudice Under the Microscope: The Implicit Association Test (Part III)

Editor’s Note: This article is part of an ongoing symposium on white fragility and its related concepts. To view all of the essays in this series, click here.

In Part II of this series, evidence for the race Implicit Association Test (IAT) was evaluated as to (1) the plausibility of the underlying construct that it claims to measure, (2) the reliability (stability) of its scores, and (3) its validity for predicting laboratory behaviors assumed to be indicative of racial prejudice. In this final installment, evidence for the test’s ability to predict noxious behaviors in the real world is discussed. Part III concludes with an overall evaluation of the race IAT’s utility for widespread use.

Category #4: Is the Race IAT effective in Predicting Noxious Attitudes and Behaviors in The ‘Real World’?

Contemporary psychology abounds with inventories, rating scales, and questionnaires designed to measure racial, sexual, and sexual orientation discrimination outside of the research laboratory. One critic of the race IAT puts the matter succinctly: ‘Although individual persons may not feel racist, the test purports to show that they will act racist in real-life settings’. Singal discusses some research studies that make sincere attempts to correlate IAT scores (race or otherwise) with ‘real-world’ behaviors. But even in these studies, ‘racism’ or ‘racial prejudice’ must be inferred from behaviors that — on surface observation — are far removed from direct observations of actual bigoted actions or the effects of such actions on the behavior of individuals (e.g., days absent from work, working more slowly on the job).

The Messiness and Difficulty of Attempts to Discern Racism in the Real World

Most readers have heard of laboratory studies in which subjects are presented with contrived data about fictitious persons, all of which have been carefully designed to portray such persons as having identical qualifications, skills, background information, or behaviors — except for names that either ‘sound black’ or ‘sound white.’ Because relevant variables can be scrupulously controlled in laboratory settings (by controlling the content of the protocols examined by research participants), researchers are on a more solid footing for interpreting racial prejudice should study participants grant less favorable evaluations/decisions to applicants with black-sounding names. This effect has been demonstrated in studies of public school teachers needing to determine punishment for discipline infractions, studies of employers needing to hire applicants for jobs advertised in help wanted ads, and studies involving written requests for information from local public services.

Unfortunately, there is a proverbial ‘elephant in the room’ that is frequently kept offstage when such findings purport to validate the simplistic morality play of ‘white wickedness/black victimhood.’ In the real world, there are persistent and stubborn racial group differences in socially significant outcomes that cannot be ignored. For example, African American students (on average) achieve the lowest average National Association for Educational Progress (NAEP) school achievement test scores compared to all other ethnic/racial groups, and this has held fairly constant throughout all decades of NAEP testing, as well as in other forms of testing. Black students display higher rates of discipline problems (on average) in integrated schools, which lead to higher rates of disciplinary sanctions (i.e., suspensions/expulsions). Among those who have completed high school or received a General Equivalency Diploma (GED), black and Hispanic students are less likely to enroll in college compared to their white peers. A 2016 report found that higher percentages of black students enroll in community colleges compared to white peers with the same high school grade point averages. Black students also attend the least selective higher education institutions with the fewest resources. Lower percentages of black students, when compared to white students, obtain bachelor’s degrees.

More African-Americans than Caucasians have poor credit scores, such that whites earning less than $25,000 per year are more likely to have better credit scores than African Americans earning between $65,000 – $75,000 per year — a finding which inevitably results in racial disparities in home mortgage lending rates. To date, roughly 70 percent of black children are born out of wedlock. In America, blacks proportionally have the highest crime rates of any racial/ethnic group — a pattern which holds true for virtually all crime categories and age groups.

Whites are expected to not even notice such facts. If these facts are noticed, they certainly are not allowed to be discussed openly — unless the discussion remains tethered to ‘politically correct’ causal interpretations (i.e., racism, discrimination, prejudice). Interpretations that attempt to explain away or mute the implications of inconvenient facts are plentiful. As one among many examples, embarrassing facts on interracial crime are breezily justified by the Southern Poverty Law Center as follows:

“So while it is true, for instance, that blacks rob whites far more than vice versa, that is hardly a surprise – whites, after all, own nearly 10 times the wealth that blacks do on average. They also own far more businesses. Thus it is only natural that any rational robber would select whites over blacks as victims . . . That poor people are more prone to criminality at the expense of the wealthy is utterly unsurprising.”

According to like-minded thinkers, therefore, only hardcore racists would believe that statistics on racial differences in socially important outcomes are ‘real’ (i.e., not attributable to white racism, bigotry, or discrimination). It comes as no surprise, therefore, why ‘racism’ is the most preferred interpretation of race IAT results that appear to disfavor blacks.

This highlights the challenges inherent in conducting ‘race research’ in real-world settings, where careful subject selection and the control of extraneous variables is extremely difficult. Conducting research in ‘real-world’ settings poses seemingly insurmountable challenges — not the least of which is the difficulty of trying to control numerous variables that work interactively to complicate or undermine the clear and unambiguous interpretation of results.

The Complexity of Interracial Social Interactions

In order to illustrate the complex psychological, moral, philosophical, and ideological variables inherent in real-life situations — as well as the insurmountable roadblocks they pose to simplistic interpretations about interracial interactions — consider the following thought experiment:

A young black boy/man asks a white girl/woman out for a date and is politely turned down. Without any further information about the actors involved or the social context of this situation, we can assuredly expect to hear accusations of ‘racism’ from the usual sources. When this situation is analyzed through a broader lens, however, the human emotions experienced by both parties are universal — regardless of whether or not racial issues play a role. That is to say, nearly all males can relate to the bruised ego and embarrassment that is felt  from being rejected, and most females can relate to the irritation of being romantically pursued by someone in whom they have little or no interest. While many psychologists and mental health counselors promote the popular narrative that ‘racism damages the mental health and psychological well-being of persons of color,’ the truth is that life itself negatively affects the mental health of all people — regardless of whether race is a factor or not.

A researcher cannot put this fictitious girl in a time machine, transport her back in time, then record her response to a white boy to see if she would have made a different decision. Researchers are therefore limited as to what they can conclude from this isolated scenario. Unknown are the numerous factors about the boy, the girl, or the social context that may complicate simplistic charges of ‘racism.’ The girl may be despondent from a broken heart caused by the recent breakup of a previous relationship (from which she still has deep feelings for her former boyfriend), and may simply have no desire to quickly jump into another relationship. Or, the boy may have approached the girl in a brutish and crude manner that offended her — which makes her rejection of his advances a ‘no-brainer.’ Or, it could be that she genuinely likes the boy and wants to go out with him, but her parents strongly disapprove of interracial dating and have threatened to disown her if she does. In this case, her rejection of the boy stems from fear of her parents and not necessarily from racial animus toward the boy. Or, it could be that the girl has dated many black boys in her past, but that she judges this particular boy as not handsome or high-status enough to waste her time with him.

Suppose that this girl truly harbors racial animus in her heart and feels personally disgusted at the thought of dating a nonwhite person. This girl would represent the kind of thinking that modern progressives feel deserves banishment from civil society. As noxious as these attitudes may be, doesn’t this girl have the right to freedom of association, guaranteed for all American citizens under the U.S. Bill of Rights? If this girl shows no evidence of ill treatment toward nonwhite minorities in other areas of her life, then what would be the point of using the race IAT to identify her as in need of ‘intervention’?

Let’s complicate this hypothetical scenario even further. Suppose that this girl has a bias in favor of black boys, where she prefers to date black boys exclusively. If IAT supporters argue that racial bias — in and of itself — is inherently bad, then why would racial bias in favor of dating black boys be any less objectionable than racial bias in opposition to dating black boys?

The Problem with Defining ‘Discrimination’

Fundamentally, there is a serious problem with the sloppy manner in which words/terms are carelessly used in social commentaries that flow from race IAT research studies. Words and concepts such as ‘disparities,’ ‘racism,’ ‘prejudice,’ ‘bias,’ ‘preferences,’ and ‘discriminatory’ are lumped together into one amorphous blob, as if they are all interchangeable in meaning. For example, consider these snippets from one article criticizing the race IAT. Here, “hot-button” words are highlighted in bold print:

“When I first took the implicit association test a few years ago, . . . [t]he test found that I had no automatic preference against white or black people . . . I was a person free of racism . . . I took the IAT again a few days later . . . Then I took the test again later on . . . I was at a loss as to what this test was telling me . . . [should I average the results] essentially showing I had no bias at all? . . . uncovering people’s subconscious implicit biases seemed like the way to show people that they really can be and are still racist . . . Project Implicit . . . [raises] awareness about implicit biases and how racism and other kinds of prejudice still exist within American society today . . . It seems like the IAT predicts some variance in discriminatory behaviors. . . .[however] it’s questionable just how useful the IAT really is for predicting biased behavior. . . I worry that an obsession by some with implicit bias . . . [may hinder the examination of] factors that are far more influential and important in shaping discriminatory behavior and that create . . . unjust ethnic disparities . . .”

In writing about a subject as visceral and as volatile as race relations, it is easy to understand how these concepts are often viewed as interchangeable in meaning (i.e., in the sense that they all connote something ‘bad’). However, this does not excuse or mitigate the confusion that results when no effort is taken to carefully distinguish between these concepts in the context of everyday living.

Consider for example the manner in which the word ‘discrimination’ is used. Race IAT research, whether conducted by its supporters or its critics, implicitly assumes that discrimination — in and of itself — is inherently ‘bad.’ In reality, however, the essential meaning of ‘discrimination’ has no inherent moral connotations. Discrimination, in its broadest sense, simply refers to ‘an ability to discern differences in the qualities of people and things, and choosing accordingly.’ Standardized tests, for example, are supposed to discriminate — in the sense that their sole purpose is to reveal individual differences in the underlying trait being measured. Otherwise, why use tests?

During the normal course of the day, persons discriminate among the relative advantages/disadvantages of waking up earlier or staying in bed for extra sleep. Persons discriminate between whether to have cereal for breakfast versus eggs and bacon — or whether to have a cinnamon roll, bagel, or English muffin. Persons discriminate among the color of socks they will wear for the day — and will eventually choose one pair to wear over another. Persons discriminate in buying cars (choosing one over another), choosing a school for their kids to attend (selecting one over another), choosing a mate to marry (and ignoring others), or choosing a neighborhood in which to live (and not another). In the broadest sense, the girl/woman in the aforementioned scenario discriminates between the implications of accepting or not accepting an invitation for a date — the outcome of which is idiosyncratic to her own unique personal circumstances. The moral evaluation of her choice depends on numerous factors that are unknown to laboratory researchers, who see only the skin color of the actors. Said differently, her decision may indeed be evaluated through a moral lens, or morality may have nothing at all to do with her decision.

Unlike laboratory settings, human beings in the real world are not randomly distributed by race with respect to the neighborhoods, schools, social clubs, or churches they attend — nor are groups randomly distributed by race with respect to social, economic, educational, artistic, or athletic accomplishments. This means that if a particular racial/ethnic group is 13 percent of the American population, we should not expect that they will be represented at 13 percent in all possible real-life settings, circumstances, or outcomes (as these percentages may be significantly higher or lower). As each human being goes through life, they most assuredly discriminate among all acquaintances that cross their path as to which persons will/will not be close friends — if they have friends at all. If persons from minority groups are relatively scarce within the pool of acquaintances that a majority-white community routinely encounters in its schools, neighborhoods, employment settings, or churches of attendance — then how can they be accused of ‘racial discrimination’ if they have little to no opportunities to interact closely with nonwhites (and vice versa for nonwhites having few chances to interact closely with whites)?

Discrimination in the Workplace

Similar problems are encountered in attempts to understand what ‘discrimination’ means within the world of business and employment law. One writer, reporting on a recent 20 million-dollar sex discrimination settlement from Walmart, warns audiences that ‘employers should do more to uncover the implicit bias preventing individuals from advancement.’ In these contexts, ‘uncovering unconscious bias’ is seen as the golden key that will unlock solutions for any problems that organizations have with respect to race issues.

In the real world, however, job advancement is contingent upon the quality of an employee’s work performance in his current job. As the late Professor Walter Williams reminds readers in his excellent book Race & Economics: How Much Can Be Blamed on Discrimination?, it is in companies’ best financial interests not to engage in racial discrimination. Nevertheless, if an employee feels that he has not been given proper advancement opportunities, companies have human resource departments and other internal grievance mechanisms (as well as redress through the courts) to utilize if they suspect racial discrimination.

Race IAT advocacy is attractive to businesses because its supposed benefits can be translated into the ‘bottom line’ — money. Businesses believe that use of the race IAT test — in addition to implementing corresponding ‘diversity training’ — would save them money by warding off discrimination lawsuits and enabling them to be more competitive. True Office Learning, for example, warns that businesses ‘cannot afford’ to live with implicit bias in their ranks if they do not want their ‘bottom line’ to be affected. In their view, the mere perception of implicit bias in the workplace causes disengagement at work to the tune of $450 to $550 billion dollars per year. They further state that unconscious bias contributes to the lack of diversity in boardrooms — which stalls performance and creates opportunities for more diverse organizations to gain a competitive advantage.

These inflated claims have no basis in reality. This comes into sharper focus when three key issues are examined: (1) the irrelevance of the ‘unconscious’ in everyday business interactions, (2) the actual legal criteria used for evaluating a business’ vulnerability to discrimination lawsuits, and (3) the apparent failure of unconscious bias/diversity training.

The Dubious Relevance of the Unconscious

Advocates use the race IAT in business/commerce as a makeshift ‘lie detector test’ for uncovering unconscious racial animus despite a persons’ conscious protestations to the contrary. If a business can identify a potentially racist employee by unearthing deep-seated sin presumably embedded in his unconscious — then requires said employee to undergo ‘diversity/sensitivity training’ — the company then gains public-relations kudos for ‘proactively doing something’ to ‘combat discrimination’ in its ranks.

In reality, however, there is absolutely no need to invoke the ‘unconscious’ in efforts to  improve business-customer relations. Hundreds of thousands of store clerks, cashiers, and salespersons across America learn to put up with (on a daily basis) difficult customers that are rude, smelly, obnoxious, demanding, or duplicitous. Employees who value their jobs know that they must resist their natural impulses and ‘go the extra mile’ to be gracious and accommodating to such customers — or else risk losing their jobs. The salesperson’s private or ‘unconscious’ feelings about the race (or any other characteristic) of customers are totally irrelevant to effective service if businesses want to attract customers, make money, and stay afloat.

The Meaning of ‘Disparate/Adverse Impact’

The term ‘disparate impact’ (also called ‘adverse impact’) is the name given to a judicial theory that allows plaintiffs to charge defendants with ‘discrimination’ in employment, business, or educational practices. The cornerstone of disparate impact theory is that employment, business, or educational practices can be ‘nondiscriminatory’ on their face, yet exert a disproportionate effect on members of legally protected groups. To illustrate, suppose a shipping company requires its loading dock employees to be able to bench press at least 50 pounds (viewed by the company as necessary for successfully loading and unloading large items for transport on trucks, boats, or planes). Since men, on average, have stronger upper body strength compared to women, this requirement for employment causes women to be disadvantaged in their probability of being hired for the job (resulting in proportionally more male applicants being hired for the job than female applicants). In legal challenges, plaintiffs must show that the defendants’ job requirements were adopted with discriminatory intent, and the defendants must show that its hiring standards are directly related to successful job performance. If the defendant (business or company) cannot demonstrate this, then it must make reasonable attempts to establish alternate hiring criteria that are directly related to successful job performance — but do not result in a disparate impact on legally protected groups.

Companies, businesses, or educational institutions that have used intelligence or scholastic aptitude tests, high school graduation requirements, specific height/weight or age requirements, or the possession of a doctoral degree as a condition for employment, have been vulnerable to accusations of racial or gender discrimination — since racial/gender groups do not show equivalent outcomes in performance on these hiring criteria. In court, plaintiffs’ success or failure in disparate impact discrimination lawsuits are contingent on additional subtleties and nuances of antidiscrimination law. The point here is that an individual’s display of personal ‘discrimination’ (i.e., of the type feared by race IAT advocates) does not have to be present for plaintiffs to win ‘discrimination’ lawsuits against businesses/organizations (based on disparate impact theory).

In 2012, African American employees brought a class action lawsuit against the executive branch of the State of Iowa, claiming racial bias in hiring and promotion within 37 departments. The plaintiffs’ claim relied on implicit bias theory. The theory’s main proponent, Dr. Anthony Greenwald, testified on behalf of the plaintiffs. According to Greenwald’s argument, approximately 70% of whites show an automatic preference for whites over blacks in race IAT studies, which (in Greenwald’s view)  necessarily influences employment decision-making absent strict bureaucratic supervision. The judge flatly rejected Greenwald’s argument using the following counterarguments as grounds for rejecting plaintiffs’ claims (paraphrased):

• Subjective discretion in hiring/promotion/firing decision-making is a reasonable method of conducting business and does not necessarily infer racial discrimination.
• (There is no evidence that) a racial preference identified by the IAT necessarily results in racially prejudicial behavior.
• (There is no evidence that) a racial preference identified by the IAT explains statistical disparities among racial groups in hiring or promotions.
• (There is no evidence that) ‘racially stereotyped thinking’ explains outcomes in discretionary employment decisions.
• No IAT data was collected on employees or managers for the State of Iowa that were relevant to the case under consideration — as it was merely assumed that the percentages found in IAT studies would be comparable for the State of Iowa.

The Failure of Unconscious (Implicit) Bias/Diversity Training

Implicit bias training has become ‘the new big thing’ in areas as diverse as social media platforms, higher education, the judicial system, and the health care professions. As Musa al-Gharbi has demonstrated, diversity training falls far short of its initial expectations — leading to deleterious outcomes despite enthusiastic (but data-lite) testimonials that promote its benefits.

For example, roughly two-thirds of all police departments across the country offer implicit bias training, and most of these make such training mandatory. The New York Police Department implemented an implicit bias training program in 2018 to reduce excessive amounts of force against African American suspects and the use of racial profiling to stop and arrest African Americans at greater rates. After the NYPD spent 5.5 million dollars on this program, researchers found that such training had absolutely no effect on officer behavior.

Even when implicit bias training is not based on laboratory IAT results, it is often a magnet for retaliatory lawsuits from plaintiffs who claim that such training is divisive, ideologically based, and disparaging of certain groups (e.g., whites, males, Christians).

IAT and the Brain/Mind Distinction

In their 2013 book ‘Brainwashed: The Seductive Appeal of Mindless Neuroscience,’ psychiatrist Salley Satel and psychologist Scott Lilienfeld wrote a trenchant critique of how results from brain imaging technology are often overstated, exaggerated, oversold, and  ‘mindlessly’ applied to problems in the legal, business, clinical, and philosophical domains of life.

Eerie parallels can be drawn between the premature enthusiasm that greeted brain imaging studies and the general public and popular media’s similar embrace (and misinterpretation) of race IAT research. In Satel and Lilienfeld’s analysis, the fundamental flaw in these movements rests in the blurring of important distinctions between the brain versus the mind.

The Brain . . .

In Greenwald et al.’s original race IAT paradigm, a computerized reaction time task involves both physiological and cognitive processes that occur within the brain between the time a stimulus (words or pictures) appears on the computer screen and the time the subject presses a computer key to indicate his response. Physiologically, the computer stimulus must first be registered by the eyes; then neural impulses travel by nerve pathways to a specific location in the brain where the stimulus is recognized as belonging to a binary category (black vs. white; pleasant vs. unpleasant). Cognitively, a decision is made (i.e., which button to push), after which a neural impulse travels from the brain to either the left or right hand, where nerves in the hand muscles respond.

Even though this has been described in multiple steps, these separate steps occur in mere fractions of a second — the exact measurement of which lies beyond a person’s conscious awareness. For example, in pressing the snooze button on one’s alarm clock after the alarm goes off in the morning, no one can provide a precise measurement of how much time was taken for the alarm clock buzzer to be perceived by the ear, the time for this information to reach the brain, the time it takes for a person to decide to press the snooze button, the time neural impulses travel from the brain to the hand, or the time it takes for the hand to make contact with the snooze button. Here, individual differences in total reaction time are a function of complex interactions between the quality of one’s hearing, neurological capabilities of the brain, sleep and arousal patterns controlled by the brain’s hypothalamus, the physical distance between one’s bed and the alarm clock, and idiosyncratic factors that influence conscious decision-making from morning to morning.

. . . Versus The Mind

In contrast to these functions that are largely associated with the brain and nervous system, the mind is infinitely more complex and is to a large extent unfathomable. As Satel and Lilienfeld argue, one cannot use the physical rules at the cellular level [of the brain] to completely predict activity at the psychological level [of the mind]. The mind consists of thoughts, contemplations, memories, imaginations, moods, feelings, fantasies, guilt, perceptions, motives, a sense of responsibility, morals, intuitions, intentions, the ability to learn, and the ability to plan for the future — all of which are powerfully shaped by individual differences in personality, previous life experiences, as well as individual differences in sociocultural forces (which can change considerably over a person’s lifetime).

Irrespective of whatever is found in the physical features of the brain, human beings are not robotic automatons in everyday life — passive ‘Manchurian candidates’ whose latent racism (however defined) is automatically activated by a face belonging to a person of a different race. The human mind possesses the ability to choose, learn, adapt, and develop  in response to a wide variety of environmental contingencies that operate in everyday life. These, in turn, shape the subsequent choices that individuals make as they progress through life. The ability to choose with the mind is what makes individuals capable of maturing in their character, being educated through learning, or being ruled/constrained by societal norms and laws (although imperfectly). How this shapes each person’s life trajectory may be shared by certain classes of individuals in some respects, but can also be unique to each individual in many other respects.

Concluding Thoughts

As Prof. Glen Loury observes, racial problems and racial conflict are perennial in America, having been present since our nation’s founding centuries ago. If we can get rid of ‘racism’ — so the thinking goes — America’s racial problems will most assuredly melt away eventually. There are three elements of this false narrative that, when combined into a combustible solution, explain social psychology’s enthusiastic (but misguided) embrace of the race IAT: (1) America’s desperation to find a ‘magic pill’ for solving racial problems that are presumed (incorrectly) to be due almost exclusively to racism (however defined); (2) the lack of basic humility in reductionist psychology, which holds fast to the idea that anything objectionable identified in the brain (or which can be explained by brain functions) must have direct and ‘mechanistic’ implications for psychological behavior in the real world that is ruled by the mind; and (3) the irresponsibility of popular print and digital media, whose stock-in-trade is the reckless promotion of ‘juicy’ stories with attention-grabbing but irresponsible headlines (e.g., ‘Is Everybody a Little Bit Racist?’; ‘You May be More Racist Than You Think, Study Says’).

In a lengthy article criticizing problems in the psychometric features and real-world applications of the race IAT, Nagai offers the most succinct summary of the essential ‘take-away’ of this essay:

Given the high probability of errors associated with the IAT, it should not be incorporated into public policies, such as hiring and university admissions, housing, banking, and government contracting, by law enforcement, in lawsuits, or in jury selection. Although it has been hailed by the media as uncovering a dark, secret side of the American psyche, numerous critics of the IAT have demonstrated that it simply cannot predict how test takers will act in the real world. The test fails to prove that we are a nation of unconscious racists.

Returning to the Star Wars analogy that opened Part I of this essay, there are no such entities as ‘race midichlorians’ that can be captured in a test tube and interpreted as reflecting an individual’s level of racial prejudice, bigotry, or propensity for racial discrimination in the real world. Whether academicians, researchers, policy makers, or the lay public can cultivate the basic humility required to truly appreciate this truism remains to be seen.

Image: Yassine Khalfalli, Public Domain