Can race be erased? Coalitional computation and social categorization

Abstract: Previous studies have established that people encode the race of each individual they encounter, and do so via computational processes that appear to be both automatic and mandatory. If true, this conclusion would be important, because categorizing others by their race is a precondition for treating them differently according to race. Here we report experiments, using unobtrusive measures, showing that categorizing individuals by race is not inevitable, and supporting an alternative hypothesis: that encoding by race is instead a reversible byproduct of cognitive machinery that evolved to detect coalitional alliances. The results show that subjects encode coalitional affiliations as a normal part of person representation. More importantly, when cues of coalitional affiliation no longer track or correspond to race, subjects markedly reduce the extent to which they categorize others by race, and indeed may cease doing so entirely. Despite a lifetime’s experience of race as a predictor of social alliance, less than 4 minutes of exposure to an alternate social world was enough to deflate the tendency to categorize by race. These results suggest that racism may be a volatile and eradicable construct that persists only so long as it is actively maintained through being linked to parallel systems of social alliance.

Based on years of research in which they were unable to prevent subjects from categorizing individuals by race, psychologists had thought that the human mind has circuits designed to automatically encode the race of each individual we encounter, as a normal part of impression formation. But the idea that automatic racial categorization would be an evolved feature of the human mind is implausible from an evolutionary point of view. Our hunter-gatherer ancestors would rarely – if ever – have encountered a person of a different race, so natural selection could not have favored brain mechanisms designed to notice and remember a non-existent dimension of ancestral social life. This line of thinking implies that race encoding is a side-effect of some other kind of computation: something that would have been useful during our evolutionary history. (Click here for discussion of the scientific status of the concept of “race”)

In PNAS #5414, we test an alternative hypothesis: that the (apparently) automatic and mandatory encoding of race is instead a byproduct of brain mechanisms that evolved for an alternative function that was a regular part of the lives of our foraging ancestors: detecting coalitions and alliances. Hunter-gatherers lived in bands, and neighboring bands frequently came into conflict with one another. Similarly, there were coalitions and alliances within bands, a pattern found in related primate species and likely to be far more ancient than the hominid lineage itself. To negotiate their social world successfully, and to anticipate the likely social consequences of alternative courses of action, our ancestors would have benefited by being equipped with neurocognitive machinery that tracked these shifting alliances.

Brain mechanisms for detecting coalitions should pick up on patterns of coordinated action, cooperation, and competition. They should also boost the saliency of any visual marker that suggests who is allied with whom. Otherwise arbitrary cues – such as skin color, accent, or manner of dress – should pick up significance only insofar as they acquire predictive validity for coalitional membership. In societies that are not completely racially integrated, shared appearance – a highly visible and always present cue – may be correlated with patterns of association, cooperation, and competition. Under these conditions, coalition detectors may perceive (or misperceive) race-based social alliances, and the mind will map “race” onto the cognitive variable coalition.

By creating an social context in which race was not predictive of a cooperative alliance, we were able to drastically decrease the extent to which subjects encoded race. We also tested the hypothesis that our minds will pick up on ANY visual marker that is correlated with patterns of cooperation and alliance, however arbitrary. The experiments showed that a new and arbitrary coalition can be encoded just as strongly as race usually is – indeed, more so – when an arbitrary yet visible cue, such as shirt color, is correlated with coalition membership. Being able to endow an invented coalition – one with no historical time depth – with the properties of “race” is what one would expect if race were encoded merely as a byproduct of coalition encoding.

Although our experiments were carried out in the laboratory, it appears that the predictions were borne out in the aftermath of September 11.

  • In our experiments, categorizing by race diminished when race did not predict the relevant coalitional conflict. Similarly, in the wake of the attacks on the World Trade Center, there were reports of less prejudice against African-Americans in New York City, and less distrust of white police officers (“us” and “them” no longer being framed as black versus white, but as Americans (blacks and whites both part of the in-group) versus Al-Qaeda terrorists); see New York Times, October 10: “Sept 11 Attack Narrows the Racial Divide”
  • In our experiments, people quickly came to use an arbitrary visual marker to predict coalitional alliances. This happened – with tragic consequences – as soon as it was suspected that Middle Eastern terrorists were responsible for the WTC attacks: hate crimes were perpetrated against individuals (e.g., Sikhs) who had coalitional cues (turbans) reminiscent of headgear worn in the Middle East.

Frequently Asked Questions:

Q: How can you tell whether someone is encoding someone’s race? (or their sex, or their coalitional alliances?)

A: By looking at systematic memory mistakes. You are asked to form impressions of individuals whom you will see engaged in a conversation (on a computer terminal). You then see a sequence of sentences, each of which is paired with a photo of the individual who said it. Afterwards, there is a surprise memory task: the sentences appear in random order, and you are asked to remember which individual said each sentence. Mistakes reveal encoding: People more readily confuse individuals whom they have encoded as members of the same category than those whom they have categorized as members of different categories. For example, a citizen of Verona who had encoded coalition membership would make more within-category errors – errors in which s/he confused, say, a Capulet with a Capulet (or a Montague with a Montague) – than between-category errors – ones in which s/he confused a Capulet with a Montague or vice versa. Similarly, a person who had categorized each individual by their race would make more within race errors – errors in which s/he confused a black person with another black person (or white with white) – than between race errors – ones in which s/he confused a black person with a white person (or white with black).

Q: Why did psychologists think that our minds categorize others by their race in an automatic and mandatory fashion (i.e., across all social contexts and with equal strength)?

A: Using this memory confusion method, they found that people always categorized the race of the individuals they saw. No matter how they changed the social context, the results did not change. While they held out the hope that something would work, they became very discouraged.

Q: In your experiments, 4 minutes of exposure was enough to reduce categorization by race. Does participating in your experiment “cure” people of racism?

A: No. Our claim is not that 4 minutes of exposure causes people to stop categorizing by race for the rest of their lives. Our hypothesis is that people are good at picking up on shifting patterns of alliance, and that this is why they can so quickly adapt to an alternate social world, one where race does not predict who is allied with whom.

Not only did 4 minutes of exposure to an alternate social world diminish categorization by race, it also caused people to start strongly categorizing by coalition membership, even though these coalitions had been invented by the experimenters. In fact, subjects categorized by coalition even more strongly than their strongest categorization by race. This was despite the fact that this was an instantly created coalition, whereas people have a lifetime of experience of patterns of association by race. In other words, long time depth is not needed for people to start encoding new coalitional alliances. This is similar to the situation after Sept 11, when people began – almost instantly – a to use cues associated with being Middle Eastern to (often mistakenly) categorize others as potential terrorists.

Q: On your theory, why do people use race to judge who is allied with whom?

A: People interact more with their neighbors and co-workers than with other people, so to the extent that a society remains even somewhat racially segregated, there will be patterns of cooperation and alliance that are correlated with race. Our coalition detectors will pick up on these patterns, and boost the saliency of any visual marker correlated with them, such as skin color. The hypothesis we tested is that our minds will pick up on ANY visual marker that is correlated with patterns of cooperation and alliance, however arbitrary.

Q: Why should visual markers matter?

A: Like other behaviors, actions that reveal coalitional dispositions are usually transitory; they are usually not available for inspection when you need to make a decision that depends on knowing someone’s coalitional affiliation. Accordingly, alliance tracking machinery should be designed to note these rare revelatory behaviors when they occur, and then use them to isolate further cues that happen to correlate with coalition but that are more continuously present and perceptually easier to assay. Such cue mapping allows one to use the behavior of some individuals to predict what others are likely to do. Because this circuitry detects correspondences between allegiance and appearance, stable dimensions of shared appearance – which may be otherwise meaningless – emerge in the cognitive system as markers of social categories. Coalitional computation increases their subsequent perceptual salience, and encodes them at higher rates. Any readily observable feature – however arbitrary – can acquire social significance and cognitive efficacy when it validly cues patterns of alliance. Ethnographically well-known examples include dress, dialect, manner, gait, family resemblance, and ethnic and coalitional badges.

Q: Do your results mean that people will not stereotype others by their race?

A: Categories are not in the world; they are in our heads. Every person can be associated with a indefinitely large number of different categories: I might be white, a woman, a professor, a mother, a daughter, a Californian, a person of Greek extraction, a Darwinian, a vitamin enthusiast, a violinist: on and on. Each category might have a stereotype associated with it: a stereotype is a set of inferences that you make about what a person is like, based on how you have categorized them. Which one would you apply to me? It depends on which category you think is relevant: on how you have classified me in a particular situation. Stereotyping comes AFTER categorization: You won’t apply your California stereotype to me unless you have already categorized me as a Californian.

To stereotype someone by their race, you must first have noted and remembered their race. Our results point to conditions under which people are less likely to notice and remember a person’s race; that is, less likely to categorize a person as “a black” or “a white”. Since categorization is a precondition for stereotyping, the results suggest that racial stereotyping will be less likely to occur in the absence of racial categorization. But technically, our results were about categorization, not about stereotyping.

Q: Will it be as easy to “erase race” in places with more racial segregation?

A: We do not know. People interact more with their neighbors and co-workers than with other people, so to the extent that a society remains racially segregated, there will be patterns of cooperation and alliance based on race, and our minds will pick up on this. In some places, these patterns may be stronger than others and, therefore, more difficult to counteract. Our experiments were conducted in California; it would be interesting to see results are different in, say, the American South (maybe it would be more difficult to “erase race” there; that is, maybe it would take more exposure to an alternate social world where race does not predict alliance).

Q: Why do you say that the results suggest greater hope for reducing racism?

A: A large body of research in psychology shows that categorizing people into groups along nearly any dimension elicits discrimination: it seems to activate an “us” versus “them” psychology. Given this finding, it would be discouraging to learn that the human mind was designed such that people cannot help categorizing others by their race. This would imply that racism is intractable. Our results indicate that mentally dividing people into different “racial groups” is not inevitable; instead, coalition, and hence race, is a volatile, dynamically updated cognitive variable, easily overwritten by new circumstances.

Q: So what can be done to “erase race” in the real world?

A: Although further research is required, the results suggest that, to the extent that people of different races are working together to pursue common goals, race might become a less salient dimension of social life. The key, we would suggest, is not merely contact, but cooperation among individuals of different races. It might seem that this is putting the effect before the cause, suggesting that cooperation might eliminate racism (instead of eliminating racism leading to cooperation), but if perceptions of cooperation drive social categorization, multi-racial cooperation might help to “erase race.”

On the scientific status of the concept of race:

“Race”, as it is usually conceived, is not a sound biological category in the world. In humans, the overwhelming preponderance of genetic variation is within population and not between populations. Moreover, the genetic variation that exists is gradual and geographically graded, rather than sharply bounded. On top of this, the racial categories that we “see” do not correspond to biologically meaningful populations in the world.

For example, if we were to go back in time and travel with Marco Polo over land to China, we probably would not have developed a concept of discrete races: although people at one end of his trip looked a bit different from those at the other end, all we would have seen along the way was continuous variation on many different dimensions. (It wasn’t until the age of exploration by ships that something like the modern construct emerged; by ship you could travel thousands of miles in a short time, without seeing any populations in between the two end points. Under those conditions, the differences between the end points would be more salient, as you would not have experienced the continuous variation in between).

For more discussion on why “race” is not a natural category of the biological world, see the following:

  1. Hirschfeld, L. (1996). Race in the Making. Cambridge, MA: MIT Press
  2. Lewontin, R. (1972) Evol. Biol. 6, 381-398.
  3. Nei, M. & Roychoudhury, A. (1993) Mol. Biol. Evol. 10, 927-943.