… Suppose, jogging alone after dark, you see a young Black male ahead of you on the running track, not attired in a jogging outfit and displaying no other information-bearing trait. Based on the statistics cited earlier, you must set the likelihood of his being a felon at 25…. On the other hand it would be rational to trust a White male under identical circumstances, since the probability of his being a felon is less than .05. Since whatever factors affect the probability of the Black attacking you—the isolation, your vulnerability—presumably affect the probability of a White attacking you as well, it remains more rational to be more fearful of the Black than of the White.4
Levin erroneously suggests that because one out of four Black men is incarcerated for commission of a felony, the statistical benchmark a person should use in judging the risk of violent assault posed by a randomly selected young Black man is 25 percent. Levin’s statistics, however, say only that one in four Black males is incarcerated for a felony, not that one in four is incarcerated for a violent felony. Only the proportion of Blacks incarcerated for violent felonies can provide any kind of benchmark for judging relative risks of violent assault by race. But the typical African American male in the criminal justice system is not a violent offender.5 Most of the increase in the number of Blacks in the criminal justice system is attributable to the “War on Drugs” and stepped-up crackdowns on drug crimes.6 In fact, the majority of arrestees for violent offenses are White.7
Assuming the woman who shot the suspected robber is a more refined Bayesian, she might frame her argument as follows. Although Blacks only make up 12 percent of the population, they are arrested for 62 percent of armed robberies.8 Therefore, the rate of robbery arrests among Blacks is approximately twelve times the rate among non-Blacks. In other words, if a defender had to make a purely race-based assessment of the risk of armed robbery, it would be approximately twelve times more probable that any given Black person is a robber than a non-Black.9 Even assuming considerable bias in police arrests, the refined Bayesian might conclude, no one can honestly say that actual rates of robbery by race are even close.
One can concede the Bayesian’s point that the rates of robbery by race are “not close” and still ask, “So what?” It is far from clear what sorts of group-based robbery rates would justify the judgment that any given member of the group presents a sufficiently high risk of robbery to be deemed “suspicious.” To make the point a different way, imagine I have two drawers, one white and the other black. Into the white drawer I pour one thousand marbles, 999 of which are green and one of which is red. Into the black drawer I also pour one thousand marbles, but this time I included twelve times the number of red ones. Thus the black drawer contains twelve red and 988 green marbles, or slightly over 1 percent red marbles. Twelve times a very small fraction may still be a very small fraction.
Now, substitute the social groups “Whites” and “Blacks” for the white and black drawers respectively, make the red marbles the members of each group arrested for violent crimes, and the problem with reading too much into the relative rates of robbery by race becomes evident. Blacks arrested for violent crimes comprised less than 1 percent of the Black population in 1994, and only 1.86 percent of the Black male population.10 Recall that even a vulgar Bayesian like Levin—who equates being incarcerated with being incarcerated for a violent crime—asserts that because the incarceration rate for White males is between 2 and 3.5 percent, “it would be rational to trust a White male” you see ahead of you while jogging alone after dark. By this Bayesian’s own logic, therefore, since Blacks arrested for violent crimes make up less than 1.9 percent of the Black male population, “it would be rational to trust a [Black] male” you ran into in the dark.
Let’s assume—perhaps erroneously—that the rates of robbery by race are in some marginal sense “statistically significant.” Thus, the Bayesian asserts that he would never employ race as the sole or even dominant risk factor in assessing someone’s dangerousness. “I merely seek to give race its correct incremental value in my calculations,” he assures us with all the aplomb of Mr. Spock. Thus, in addition to race, he carefully weighs other personal characteristics—such as youth, gender, dress, posture, body movement, and apparent educational level—before deciding how to respond. Having tallied up these “objective” indices of criminality, the Intelligent Bayesian argues that his conduct was reasonable (and thus not morally blameworthy) because it was “rational.”
A threshold problem with the Bayesian’s profession of pristine rationality concerns the “scrambled eggs” problem described earlier—that is, the practical impossibility of unscrambling the rational and irrational sources of racial fears. For countless Americans, fears of Black violence stem from, among other things, the complex interaction of cultural stereotypes, racial antagonisms, and unremitting overrepresentations of Black violence in the mass media. As for the mass media, especially television news, recall the letter in the Introduction from the would-be Bayesian who remarked, “If I saw Blacks in my neighborhood I would be on the lookout, and for a good reason.” The “good reason” he cites for his hypervigilance about Blacks is television. Few Americans keep copies of FBI Uniform Crime Reports by their bedsides: when asked in a Los Angeles Times survey (February 13, 1994) from where they got their information about crime, 65 percent of respondents said they learned about it from the mass media. But television journalism on crime and violence has been proven to reveal, and project, a consistent racial bias.11
Even if media reporting on crime and violence were not biased, our minds simply do not process information about Blacks and other stereotyped groups the way the Bayesian assumes. The Bayesian assumes that our minds can passively mirror the world around us, that they can operate like calculators, and that social stereotypes can be represented in our minds as mere bits of statistical information, as malleable and subject to ongoing revision as the batting averages of active majorleague baseball players. Each of these assumptions flies in the face of what modern psychology reveals about the workings of the human mind.
As is described in detail in chapter 6, social stereotypes are not mere bits of statistical information but rather well-learned sets of associations among groups and traits established in children’s memories before they reach the age of judgment. And once a stereotype becomes entrenched in our memory, it takes on a life of its own. Case studies have demonstrated that once an individual internalizes a cultural stereotype, she unconsciously interprets experiences to be consistent with the underlying stereotype, selectively assimilating facts that validate the stereotype while disregarding those that do not.12 The tendency of individuals to reject or ignore evidence that conflicts with their cultural stereotypes expresses itself in many forms, perhaps none as perplexing as the backhanded “compliment” some White liberals think appropriate to bestow on “deserving” Blacks: “I don’t think of you as Black.” For Blacks who harbor the hope that their personal achievements can “uplift the race” by upending stereotypes, these clumsy bouquets are deeply disturbing. The more success you achieve, the less likely that your success will redound to the reputational benefit of your community. In the words of Evelyn Lewis, the first Black woman to make partner in a major San Francisco law firm, “[W]hat you do well will reflect well on you, but only as an individual. And what you do poorly—well, that’s when what you do will be dumped on the whole race.”13 To the extent that the Bayesian aggressively assimilates negative statistical information about Blacks while remaining oblivious to contradictory or positive statistical information, she undermines her claim of objectivity.
Further, the Bayesian’s contention that she can delicately balance the racial factor in her calculations is refuted by recent discoveries about the psychological impact of stereotypes. A stereotype, unlike ordinary statistical information, radically alters our mindset, unconsciously bringing about a sea change in our perceptual readiness. Under the influence of a stereotype, we tend to see what the stereotype primes us to see. If violence is part of the stereotype, we are primed to construe ambiguous behavior as evincing violence, not on