Kristina Olson
Social Psychology Program
Harvard Graduate School of Arts and Sciences, Cambridge, MA 02138

Hidden Bias: Talk Synopsis

Kristina Olson
(and Project Implicit including Mahzarin Banaji, Tony Greenwald, and Brian Nosek).

For decades psychologists have been interested in the idea that we might have thoughts, attitudes or other cognitions, in our minds that we have limited or even no access to (e.g., Freud, 1900/1913). Today I'm going to talk a little about a new measurement technique that our research group has been developing that examines one particular kind of non-conscious or what we call “implicit” cognition, social group attitudes and beliefs. I'm going to suggest that we have social attitudes and beliefs that we don't have full introspective access to and that these attitudes (that I'll call implicit attitudes) not only exist but may even predict our behavior better than the attitudes you can tell me about (what I'll call explicit attitudes).

The Implicit Association Test

About 10 years ago our lab designed a task called the Implicit Association Test (IAT) which is now available for anyone to take on the web at implicit.harvard.edu. This test measures the associations people hold between concepts. Previous research has demonstrated that people are quicker to associate items that are related (e.g., “doctor” and “nurse”) compared to items that are unrelated (e.g., “doctor” and “screwdriver”) (Neely, 1976). Our research team asked whether this information could be used to create a test measuring the associations people have between particular social groups and stereotypes or overall goodness and badness. Researchers in our group tested this possibility by asking whether, for example, people are faster to associate women with humanities and men with science compared to their speed of associating men with humanities and women with science.
 [visit implicit.harvard.edu à demonstration site for an example of the IAT and get your score!]

What began as an interesting test, has since resulted in more than 200 studies and more than 2,000,000 completed online tests (Nosek et al. 2007). These studies have demonstrated, for example, that the majority of respondents are quicker to associate White people with good things (e.g., words like “terrific,” “great”) and Black people with bad things (e.g., words like “terrible,” “awful”) compared to White with bad things and Black with good things (what we'd call an implicit preference for White over Black), and similarly the majority of respondents show a preference for Young people over Old people , Rich people over Poor people, Thin people over Fat people and Straight people over Gay people on the IAT (Nosek et al., 2007). Additionally, this method has been used to test people's implicit stereotypes, for example, showing that people are quicker to associate women with family and men with career compared to women with career and men with family (Nosek et al., 2007).

One particularly interesting finding is that while people often report that they don't prefer Whites to Blacks when we ask them, they still will often show this preference implicitly as demonstrated by the IAT (Greenwald, McGhee, & Schwartz, 1998). Overall, while we find a correlation between implicit attitudes (as measured by the IAT) and explicit attitudes (when we ask people how they feel), this correlation is a small one and is only generally seen when we have very large samples of participants. Additionally, at the mean level, implicit and explicit attitudes diverge considerably. For example, in a lab study, Cunningham, Nezlek, and Banaji (2000) found their sample on average showed no preference for White people over Black people or Rich people over Poor people and only a slight preference for Straight people over Gay people when asked explicitly. In contrast, on the IAT they showed a huge preference for Black people over White people, Rich people over Poor people, and Straight people over Gay people.

Attitudes have been demonstrated to differ a bit depending on demographic variables such as gender or race. For example, on the Black-White IAT, White participants on average show a strong preference for White people over Black people whereas Black participants show no significant preference for White people or Black people (Nosek, et al., 2007). Similarly straight people generally associate Straight with good and Gay with bad whereas gay participants show no implicit preference for straight or gay (Olson, et al., 2007).

Another recently developed test is the disability IAT. This test examines how quickly people can associate the terms “disabled” and “abled” with good and bad attributes. Overall people are quicker to associate disabled with bad and abled with good compared to the opposite pairings. This preference for the abled is more pronounced for participants who self-identify as more politically conservative, for male participants and for older participants (Nosek et al., 2007).

Predictive Validity

     I have demonstrated that across several different tests we see participants associating some concepts with good and others with bad, more or less quickly-but who cares about a 200 millisecond difference? What does this 200 ms difference even mean? In the past few years researchers have become increasingly interested in whether or not the IAT predicts behavior and if it does, when it does. While many studies have been conducted, I will summarize only a few results here.

     In a study by Carney and colleagues (Carney, Olson, Banaji, & Mendes, in preparation), researchers found that the Black-White IAT predicted participants' automatic smiling behavior to faces of subliminally presented Black and White faces. That is, people who showed the largest preference for Whites over Blacks on the IAT also smiled the most to White faces relative to Black faces as measured by facial EMG. In other studies, researchers have demonstrated, for example, performance on the Black-White IAT is related to participants' social interactions with Black and White experimenters (McConnell & Leibold, 2001) and  IAT preferences for consumer products predict participants' choices of products under time pressure (Friese, Wanke, & Plessner, 2006).

     A recent meta-analysis was conducted examining the relationship between implicit and explicit attitudes and behavior (Poehlman, Uhlmann, Greenwald, & Banaji, under review). One of the central questions was whether implicit or explicit attitudes better predict behavior. It turns out the answer was something like “it depends.” The authors found that the domain of study was crucial. When researchers were interested in questions like product preference (e.g., Coke vs. Pepsi), simply asking people for their preference better predicted their behavior. This was true in almost every case. The one exception was in the domain of social group attitudes. In this case participants' implicit attitudes better predicted their behavior, suggesting that the IAT remains a useful measure for behavioral prediction within the domain of social group attitudes.

How do we Change our Implicit Attitudes and Stereotypes?

     Now that I've demonstrated that implicit attitudes exist and that they predict actual behavior and that perhaps all of us hold some implicit attitudes that we'd prefer we didn't, a likely question is “what do we do about it?” This is a relatively new research question. Buju Dasgupta and her colleagues have taken on this question in a series of studies. In her first study on the topic, she exposed one group of participants to counter-stereotypic people including Black people known for good actions (e.g., Martin Luther King, Jr.) and White people known for bad actions (e.g., Timothy McVeigh) and one group of participants were not shown these people (Dasgupta & Greenwald, 2001). Participants were given the Black-White IAT immediately after exposure and 24 hours later and the authors found that the group who had been exposed to the counter-stereotypic people showed less anti-Black, pro-White bias than the control group both immediately and 24 hours later.

     In a second paper focused on changing implicit stereotypes, Dasgupta and Asgari (2004) examined gender stereotypes in groups of college students who either attended a women's college or a co-educational college. They found that while the two groups only differed a little in their implicit stereotypes upon entrance to college (with the women's college students showing somewhat less gender stereotyping), by the next year the students at the women's college showed no significant gender stereotyping whereas students at the co-ed college showed even stronger gender stereotypes than upon entrance to college. Even more interestingly, using mediational analyses, they discovered that this shift was almost exclusively attributable to the proportion of female faculty a given participant had during her first year of college. That is, it wasn't the fact that the students attended a women's college that caused the change in attitudes, rather that the women's college had more female professors and this actually led to the decrease in gender stereotyping. Said another way, if two students, one at the women's college and one at the co-ed college had the same proportion of female professors, this result would suggest that they would have similar effects on their initial implicit stereotypes.

Summary

     In this talk I've demonstrated what the IAT is, how it works, what it predicts and some initial ways that people have changed implicit attitudes. While this technique is relatively new, more and more research is continuing to come out every year, with new extensions into other fields including clinical psychology (Teachman & Woody, 2003; Yan & Guo-liang, 2004), law (Kang, 2005), and business (Maison, Greenwald, & Bruin, 2004).

____

References

Carney, D., Olson, K. R., Banaji, M. R., & Mendes, W. B. (2007). The behavioral regulation of unconscious bias. Manuscript in preparation.

Cunningham, W. A., Nezlek, J. B., & Banaji, M. R. (2004). Implicit and explicit ethnocentrism: Revisiting the ideologies of prejudice. Personality and Social      Psychology Bulletin, 30, 1332-1346.

Dasgupta, N., & Asgari, S. (2004). Seeing is believing: Exposure to counterstereotypic women leaders and its effect on the malleability of automatic gender stereotyping. Journal of Experimental Social Psychology, 40, 642-658.

Dasgupta, N., & Greenwald, A. G. (2001). On the malleability of automatic attitudes: Combating automatic prejudice with images of admired and disliked individuals.      Journal of Personality and Social Psychology, 81, 800-814.

Freud, S. (1900/1913). The Interpretation of Dreams. Translation by A. A. Brill. New York: Macmillan.

Friese, M., Wanke, M., & Plessner, H. (2006). Implicit consumer preferences and their influence on product choice. Psychology & Marketing, 23, 727-740.

Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of      Personality and Social Psychology, 74, 1464-1480.

Kang, J. (2005). Trojan horses of race. Harvard Law Review, 118, 1489-1593.

Maison, D., Greenwald, A. G., & Bruin, R. H. (2004). Predictive validity of the Implicit Association Test in studies of brands, consumer attitudes and behavior. Journal of Consumer Behavior, 14, 405-415.

McConnell, A. R., & Leibold, J. M. (2001). Relations among the Implicit Association Test, discriminatory behavior, and explicit measures of racial attitudes. Journal of  Experimental Social Psychology, 37, 435-442.

Neely, J. H. (1976). Semantic priming and retrieval from lexical memory: Evidence for facilitatory and inhibitory processes. Memory & Cognition, 4, 648-654.

Nosek, B. A., Smyth, F. L., Hansen, J. J., Devos, T., Lindner, N. M., Ranganath, K. A.,  Smith, C. T., Olson, K. R., Chugh, D., Greenwald, A. G., & Banaji, M. R. (under      review). Pervasiveness and Correlates of Implicit Attitudes and Stereotypes.  

Olson, K.R., Parsons, A., Rowatt, W., Nosek, B. A., Mangu-Ward, K., & Banaji, M. R.  (2007).  Americans' Attitudes toward Gays and Lesbians at the Dawn of the      Twenty-first Century: Evidence from Lab and Web samples. Manuscript in  preparation.
Poehlman, T. A., Uhlmann, E., Greenwald, A. G., & Banaji, M. R. (under review). Understanding and using the Implicit Association Test: III. Meta-analysis of      predictive validity.

Teachman, B., A., & Woody, S. R. (2003). Automatic processing in spider phobia: Implicit fear associations over the course of treatment. Journal of Abnormal      Psychology, 112, 100-109.

Yan, D., & Guo-liang, Y. (2004). The Implicit Association Test and its application in      clinical psychology. Chinese Journal of Clinical Psychology, 12, 432-434.