Cesario points out that the conditions in social psychological experiments that foster the manifestation of bias are largely absent in “real-world” domains. Nevertheless, the bias detected in such flawed experiments has been deemed “widespread” and “pervasive” in “real-world” domains. The serious flaws Cesario has identified have not impeded the overselling of this “bias.” In fact, they have facilitated this overselling, some of whose costs I will now enumerate.

First, as Cesario points out, base rates of criminality, school rule violations, and other important data are not included in the stimulus materials used in social psychology experiments. People, therefore, cannot use these data, and resulting judgments necessarily don't correspond to the judgments made in the “real world” where such information is often available. When people are given the opportunity to use base rates in social psychological experiments on race, they are deemed to be “Bayesian bigots” (Banaji, Reference Banaji2003) and are perceived by observers as “unintelligent” reasoners (Cao, Kleiman-Weiner, & Banaji, Reference Cao, Kleiman-Weiner and Banaji2019). Should people ignore base rates in their inter-racial social judgments? The underutilization of base rates in criminal investigations has been shown to cost approximately 1,900 lives of minority members per year (Farmer & Terrell, Reference Farmer and Terrell2001). So which is best: (a) ignoring base rates, policing every racial group precisely equally, and annually losing minority lives to homicide, or (b) honoring base rates with more police in minority as opposed to majority neighborhoods and saving lives? Deeming Bayesian reasoners to be bigots has a serious societal cost. In the judgment and decision-making literature not using relevant base rates is considered to be an error in reasoning (Kahneman & Tversky, Reference Kahneman and Tversky1973). In the social psychology literature, not providing relevant base rates for people to use in their judgments is considered to be appropriate.

Second, as Cesario points out, the shooter bias studies typically provide neither history of nor prior interaction with the suspect, and such studies often use laypersons rather than trained professionals. Manifestations of bias in such unrealistic situations help to fuel calls to defund the police who are accused of harboring implicit bias in realistic situations. Recent analyses reveal neither racial differences in the use of extreme force on the part of police (Fryer, Reference Fryer2019) nor disproportionate arrests of Blacks (Beck, Reference Beck2021). These results contrast sharply with the results of shooter bias studies using non-professionals in unrealistic situations. A recent poll in minority neighborhoods found that over 80% of residents in those neighborhoods want either an increase in police presence or no decrease in police presence (Grzeszczak, Reference Grzeszczak2020). Police reticence, resignations, and retirements have resulted in increased crime in many of these neighborhoods during the last few years (e.g., Lauritien, Reference Lauritien2021). Debilitating the police because of their supposed “widespread” and “pervasive” bias has a serious cost.

Third, the conclusions drawn from social psychological research have prompted charges of racism even when the empirical evidence provides no support for such accusations. To cite one example, Sowell (Reference Sowell2019, p. 89) points out that in 2000 the U.S. Civil Rights Commission, on which I was a state advisory committee member, reported that 44.6% of Black applicants were denied a mortgage, but only 22.3% of Whites were denied. This led to a chorus of demands that the government should crack down on this level of abject discrimination. Cesario discussed several laboratory studies in the business domain in which discrimination was the purported motivation for such racial differences. However, further inspection of the mortgage data showed that the credit scores for the Black applicants were lower than the credit scores for Whites. Because the lending sources had “skin in the game” – the money that they would loan to the mortgage applicants – they were prudent to base their mortgage decisions on the financial qualifications of the applicant. This example illustrates the role of incentives. In social psychology laboratory studies, there is no incentive for either discriminatory or non-discriminatory behavior. In contrast, a bigoted mortgage lender who refuses to lend money to qualified Black applicants will suffer negative financial consequences. In the “real world” prejudice can have a cost. Thus, it will be less likely to be manifested than in a social psychology experiment in which bigotry goes unpunished.

Cesario points out that accusations of prejudice may occur in such instances in which a teacher might interpret a child's behavior differently depending on the race of the child. Unfortunately, there is no perfectly objective way of categorizing the child's behavior. The police face this ambiguous situation frequently in which mis-categorizing a behavior can have disastrous consequences. Fortunately, there are some instances in which the categorization of the behavior is certain and thus where prejudice can be accurately discerned. An example occurred in New Jersey when the police were criticized for giving more speeding tickets to African American drivers than White drivers. A research project was initiated in which a camera took pictures of drivers on the New Jersey turnpike while a radar gun measured each driver's speed. A trio of three persons looked at a still photograph of each car's driver, and two of the three evaluators had to agree on what the race of the driver was. Because the trio was looking at a still photograph, they could not determine if this car was exceeding the speed limit. “Speeding” was defined as traveling at least 15 miles per hour over the speed limit. The data showed that the drivers identified as African American were nearly twice as likely to be speeding as the drivers identified as White. In other words, the differential number of speeding tickets was not because of “implicit prejudice” (Hinnant, Reference Hinnant2002). Of course, laboratory experiments are unlikely to contain spontaneous levels of natural behavior which can be categorized with certainty. Thus, it is risky to generalize racial differences detected in laboratory experiments to societal-level racial differences.

Social psychology experiments can be valuable and informative. However, their generalization to societal-level issues must be done with humility, better design, and much more attention to external validity.