- White people don’t think they are more human than Blacks when asked, but implicit measures suggest otherwise.
- Problems with the Implicit Association Test (IAT) weaken the claim that Whites see Black people as less human.
- Racism is a serious problem, but Project Implicit states that the IAT cannot definitively reveal it.
A recent study has made some strong claims about an unconscious racial bias among White people (Morehouse, Maddox, and Banaji, May 22, 2023). The authors wrote that the results “highlight the tendency among socially dominant groups to reserve the quality Human for their own kind.” In an interview, Morehouse referred to “subtle ways where we seem to be stripping the humanity of people” (Elbein, 2023).
The research included over 60,000 participants and appeared meticulously conducted, and some of the conclusions appear quite serious. Reserving the quality “human” for White people and stripping humanity from others even sounds a bit barbaric. Fortunately, however, careful examination of the results and of the instrument behind them suggests that we should interpret the findings much more cautiously.
In self-report measures, White participants considered Blacks just as human, and just as not-animal, as White participants considered Whites. In fact, White participants considered Blacks slightly more human or slightly less animal than Whites (reported in a footnote).
The crucial results came from the Implicit Association Test (IAT), a measure popularly believed to reveal unconscious prejudice. Although the calculations for the IAT are a bit complex, the basic result was that White participants associated human terms more easily with Whites than with Blacks.
Specifically, White participants were slightly faster when directed to press the same key on a keyboard for both White faces and human terms (or both Black faces and animal terms), as compared to when directed to press the same key for both Black faces and human terms (or both White faces and animal terms). Black participants did not show a difference. Human terms included “human,” “person,” “man,” and “woman.” Animal terms included words like “dog,” “pig,” “snake,” and “tiger.”
Although it cannot be determined from the results exactly how much faster Whites pressed the relevant key, other IAT studies have shown significant results when the difference was only a tenth of a second. In other words, Whites probably very quickly associated human terms with Blacks, and Whites associated human terms with Whites a fraction of a second faster.
Although the difference was statistically significant across multiple studies, deserving of publication, it also deserves emphasis that the results did not show Whites discounting Blacks’ humanity. Not even unconsciously. Comparative statements are not about absolute amounts. Being “faster” at one task than another does not mean you’re slow or you failed at the other task—you can be fast at both. Thinking one group is “more” human than another does not mean you “reserve the quality Human” for the first group, especially when the “thinking” is measured by the IAT.
The IAT Controversy
Many researchers have raised serious concerns about the IAT (Arkes and Tetlock, 2004; Jussim, 2022; Jussim et al., in press; Mitchell and Tetlock, 2017; Oswald et al., 2015; Schimmack, 2021). Whether the IAT actually measures unconscious racial bias is one of many issues. In general, inferring feelings or prejudices from external behavior alone runs the risk of the fundamental attribution error (Stalder, 2018), especially when there are situational or cultural factors contributing to that behavior as they contribute to IAT key-pressing (Gawronski, 2019; Payne et al., 2017).
Schimmack (2021) wrote that “IATs are widely used without psychometric evidence of construct or predictive validity.” Jussim (2022) wrote that “most of the most dramatic claims about [the IAT] have been debunked or, at least, shown to be dubious and scientifically controversial.” Jussim has provided a public archive of about 50 articles that document the problems and limitations.
Those who support the IAT emphasize that IAT scores predict discriminatory behavior. The extent of such prediction is highly debated, and Morehouse and coauthors acknowledged they didn’t study behavioral outcomes. But the bigger picture is that even if IAT scores reflected unconscious bias and strongly predicted discrimination, the research is correlational, and correlation does not imply causation. So we still wouldn’t know if IAT-based bias can “cause” discriminatory behavior (Mitchell and Tetlock, 2017).
I don’t mean to cherry pick. There are persuasive quotes, big names, and studies that support the IAT. Morehouse and colleagues’ large-scale investigation cannot fairly be reduced to an overstated conclusion in an interview (Ahmed, 2023; Elbein, 2023). The main problem is that those who promote the IAT as a valid measure of prejudice rarely cite the other side. Science works less well when one side won’t even acknowledge the other’s existence. Psychology textbooks also typically fail to mention the problems with the IAT (Bartels and Schoenrade, 2022).
The Project Implicit website (through which data were collected for the May 22 paper) strangely provides a more balanced picture. Those behind the project include the IAT creator, Anthony Greenwald, and the senior author of the May 22 paper, Mahzarin Banaji. The Harvard website states that they “make no claim for the validity of [their] suggested interpretations” and that “the IAT cannot indicate whether a person is or is not prejudiced.” Although curious readers should visit the website to read those quotes in context, these acknowledgments undercut most IAT headlines and stand in stark contrast to some of the stronger claims made by these same researchers in other venues.
Racism and white supremacy are problems in the United States. But the IAT cannot definitively reveal personal biases that might drive those problems. Morehouse and colleagues’ aggregated data may reflect common exposure to or knowledge of, but not necessarily agreement with, the racial stereotypes that permeate our society and depict non-Whites as less human.
So if you score poorly on an IAT, you may or may not be an unconscious racist, but you may be. Individuals can reflect on that possibility for themselves. Most researchers who raise concerns about the IAT see some value in the IAT research, but they seem to want journalists and other researchers at least to be more cautious or fair-minded in stating conclusions. Conflicts between anti-racists and racism deniers are challenging enough without exaggerating the already very serious findings on racism and inequalities.
Issam Ahmed, “Harvard Study Finds Implicit Racial Bias Highest Among White People,” Yahoo News, May 22, 2023, https://news.yahoo.com/harvard-study-finds-implicit-racial-202525955.html.
Hal R. Arkes and Philip E. Tetlock, “Attributions of Implicit Prejudice, or ‘Would Jesse Jackson Fail the Implicit Association Test?’” Psychological Inquiry 15 (2004): 257–78.
Jared M. Bartels and Patricia Schoenrade, “The Implicit Association Test in Introductory Psychology Textbooks: Blind Spot for Controversy,” Psychology Learning and Teaching 21 (2022): 113–25.
Saul Elbein, “Unconscious Racial Bias Goes Deep—Regardless of Views on Equality: Study,” The Hill, May 22, 2023, https://thehill.com/homenews/4015720-unconscious-racial-bias-goes-deep-regardless-of-views-on-equality-study/.
Bertram Gawronski, “Six Lessons for a Cogent Science of Implicit Bias and Its Criticism,” Perspectives on Psychological Science 14 (2019): 574–95.
Lee Jussim, “12 Reasons to Be Skeptical of Common Claims About Implicit Bias,” Psychology Today, March 28, 2022, https://www.psychologytoday.com/us/blog/rabble-rouser/202203/12-reasons-be-skeptical-common-claims-about-implicit-bias.
Lee Jussim et al., “IAT Scores, Racial Gaps, and Scientific Gaps,” in The Future of Research on Implicit Bias, eds. Jon A Krosnick et al. (New York: Cambridge University Press, in press), https://mfr.osf.io/render?url=https://osf.io/mpdx5/?direct%26mode=render%26action=download%26mode=render.
Gregory Mitchell and Philip E. Tetlock, “Popularity As a Poor Proxy for Utility: The Case of Implicit Prejudice,” in Psychological Science Under Scrutiny: Recent Challenges and Proposed Solutions, eds. Scott O. Lilienfeld and Irwin D. Waldman (West Sussex, UK: Wiley, 2017), 164–95.
Kirsten N. Morehouse et al., “All Human Social Groups Are Human, but Some Are More Human Than Others: A Comprehensive Investigation of the Implicit Association of ‘Human’ to US Racial/Ethnic Groups,” Proceedings of the National Academy of Sciences 120 (May 22, 2023), advance online publication, https://doi.org/10.1073/pnas.2300995120.
Frederick L. Oswald et al., “Using the IAT to Predict Ethnic and Racial Discrimination: Small Effect Sizes of Unknown Societal Significance,” Journal of Personality and Social Psychology 108 (2015): 562–71.
B. Keith Payne et al., “Flipping the Script on Implicit Bias Research with the Bias of Crowds,” Psychological Inquiry 28 (2017): 306–11.
Ulrich Schimmack, “Invalid Claims About the Validity of Implicit Association Tests by Prisoners of the Implicit Social-Cognition Paradigm,” Perspectives on Psychological Science 16 (2021): 435–42.
Daniel R. Stalder, The Power of Context: How to Manage Our Bias and Improve Our Understanding of Others (Amherst, NY: Prometheus Books, 2018).