Pages

Sunday, March 30, 2014

Contest to Reduce Implicit Racial Bias Shows Empathy and Perspective-Taking Don't Work



NCAA college basketball isn't the only hot competition involving a team from the University of Virginia.  UVa Psychology Professor Brian Nosek is one of three founders of Project Implicit, a collaborative nonprofit dedicated to the study of implicit social cognition — how unconscious thoughts and feelings can influence attitudes and behavior.

Prof Nosek is also heavily involved in the Open Science and Replication movements. Along with graduate student Calvin Lai, he led a multinational group of 22 other researchers in a competition to see who could devise the best intervention to reduce racial bias scores on a widely administered implicit test, the race IAT (Lai et al, 2014).

The Implicit Association Test (IAT) is a mainstay of social psychology research that assesses implicit (unconscious) attitudes towards outgroups (based on race, sexual orientation, body size, age, etc.), stereotypes (e.g., men are in science, women are in arts/humanities), opposing ideologies (e.g, Democrat vs. Republican), and a staggering array of other binary preferences (Classical-Hip hop IAT, Astrology-Science, Britney Spears-50 cent, Boxers-Briefs, Harry Potter-Lord of the Rings and on and on). 

Or does it... ? There have been some vocal critics of the IAT over the years who have questioned what the test actually measures. I'll return to this point later, but for now let's look at the impressive aspects of the new paper.

Performance on the Black-White IAT was compared after 17 brief interventions aimed at changing pro-White bias (and a "faking" condition) relative to a control condition of no pre-test intervention. Participants were over 20,000 non-Black individuals registered at the Project Implicit website, randomized into groups of 300-400. Most of the interventions were tested in four different studies. The contest rules allowed changes to the design between studies. The goal was to lower pro-White bias scores to the point of no preference between Blacks and Whites.

In the IAT, participants classify faces as Black or White and words as good or bad. Some blocks contain only faces or only words. The two critical conditions are shown in the figure above. The stimulus-response mappings are rotated in different blocks to either reinforce stereotypes (bottom) or go against stereotype (top). In the Stereotype condition, participants press the same key when they see White faces or “good” words. They press the other key when they see Black faces or “bad” words. Most White participants (and many African Americans) show a pro-White “preference” or bias, with faster responses when White/good and Black/bad are mapped to the same key (than vice versa).

Conversely, in the Against Stereotype condition, Black faces and positive words are mapped to one key, and White faces and negative words are mapped to the other key. In essence, this induces a response conflict similar to that seen in many classic cognitive psychology tasks such as the color-word Stroop task, e.g. BLUE (say “red”) and the Eriksen flanker task, e.g. ← ← → ← ← (press right button). Slower response times in the IAT conflict conditions has been interpreted as an implicit bias against Black people (Greenwald et al., 2009), although one could argue that executive control abilities play a role here, just as they do in the Stroop task (Siegel et al, 2012).1


The Interventions

The interventions were divided into six different descriptive categories. Although the descriptions were based on existing hypotheses in the literature, they do not imply the operation of any specific psychological mechanism. The interventions had to be brief in length (5 min or less), yield interpretable scores, and have a low attrition rate. See Appendix 1 at the end of this post for a detailed list.

(1) Engage with others’ perspectives: imagine the thoughts, feelings, and actions of Black individuals (Interventions #1–3).

(2) Exposure to counterstereotypical exemplars: assigned to fictional groups with positive Black ingroup members and/or negative White outgroup members; OR think about famous Black people and infamous White people (Interventions #4–8).

(3) Appeals to egalitarian values: activate egalitarian goals (e.g., thinking about failures to be objective or egalitarian); OR think about multicultural values (Interventions #9–13).

(4) Evaluative conditioning: strengthen counterstereotypical associations by pairing White faces with Bad words and Black faces with Good words (Interventions #14 and #15).

(5) Inducing emotion: the positive emotion of elevation (Intervention #16).

(6) Intentional strategies to overcome biases: provide strategies to override or suppress the influence of automatic biases, rather than trying to shift associations directly (Interventions #17 and #18).


To reveal my own a priori biases regarding these descriptive categories, I favor (6) Intentional strategies to overcome biases, which I have written about previously (in 2008). These were interventions #17 Using Implementation Intentions, and #18 Faking the IAT as proposed by Calvin K. Lai, the first author of the manuscript.

Results indicated that nine of the interventions were effective, and nine were ineffective. The interventions that tried to change attitudes (Appeals to egalitarian values), increase empathy or perspective-taking (Engage with others’ perspectives), or elicit an elevated sense of morality (Inducing emotion - Haidt) were completely ineffective.

I note here that the failed interventions all tried to challenge the racially biased attitudes and prejudice purportedly measured by the IAT. These interventions are below the red line in the figure below.

- click on image for a larger view -


Figure 1 (modified from Lai et al, 2014). Effectiveness of interventions on implicit racial preferences, organized from most effective to least effective. Cohen’s d = reduction in implicit preferences relative to control; White circles = the meta-analytic mean effect size; Black circles = individual study effect sizes; Lines = 95% confidence intervals around meta-analytic mean effect sizes. IAT = Implict Association Test; GNAT = go/no-go association task.


Some of the most effective interventions showed variability across studies, because the parameters were altered between studies (which was allowed). Importantly, some of the interventions included multiple manipulations. The top three, Vivid Counterstereotypic Scenario, Practicing an IAT With Counterstereotypical Exemplars, and Shifting Group Boundaries Through Competition all employed Implementation Intentions in addition to the primary intervention.

What are Implementation Intentions? [in brief, think “Black = good”]
The mechanism connects an environmental cue with the goal intention, making associations between the behavior and the cue more accessible in memory. ... The task gave participants a short tutorial on how to take the IAT and informed them about the tendency for people to exhibit an implicit preference for Whites compared with Blacks. Participants were then asked to commit themselves to an implementation intention by saying to themselves silently, “I definitely want to respond to the Black face by thinking ‘good.’” 

On its own, this manipulation was effective in reducing bias scores (p = .032, d = .19). The effect size was enhanced by allowing participants to practice the task before the instructions were given (p = .00037, d = .32). In other words, once subjects were even superficially familiar with the task, being told to think “Black = good” significantly reduced pro-white sentiment (i.e., IAT scores).

This intervention is remarkably similar to my previous anecdotal findings (n=1) for the Human or Alien? test and the Dead or Alive? test. My 2008 results are below. I showed similar effects for the Black-White test and the Women in Science test, but I couldn't find the results for those.


The Neurocritic is Human AND Alien. Coming soon: “Tips for Manipulating the IAT.”

You have completed the study.
Your Result

Your data suggest little to no automatic identification with Human compared to Alien.

If your results, provided above, indicate a stronger identity with alien relative to human, then you are probably an alien.


The Neurocritic is NEITHER Dead NOR Alive. Or both Dead AND Alive. Plus, as promised, today we'll cover “Tips for Manipulating the IAT.”

You have completed the study.
Your Result

Your data suggest little to no automatic identification with Alive compared to Dead.

Your results, summarized above, are an implicit indicator of whether you are alive or dead. Implicit measures are superior to self-report because the latter is notoriously unreliable. People may report being alive because social pressures suggest that it is more desirable to be alive. Also, people may not have introspective access to their animate-status, making such self-report untrustworthy.

Super Secret Tip for Manipulating the IAT
My “faking” strategy was simple, and relied on neither deliberate slowing of response times nor a long-standing affiliation with aliens. When SELF and ALIEN were mapped to the same key, I merely said to myself, “I'm an alien.” This strategy was transient, applied only when those stimulus-response mappings were the same, not when SELF and ALIEN were mapped to different keys. I used the same strategy for the Dead or Alive IAT. In both cases, I responded as quickly and as accurately as possible.

Here, what I'm calling “faking” is the Using Implementation Intentions instructions (and not the Faking the IAT intervention of Lai et al, 2014). Again, the top three contest winners combined this strategic feature with another manipulation, as noted by the authors:
The three most effective interventions appear to leverage multiple mechanisms to increase their impact on implicit preferences...  The most effective intervention, Vivid Counterstereotypic Scenario, involved the participant as the subject of the story, had the participant imagine his- or herself under a highly threatening life-or-death situation, exposed participants to counterstereotypical exemplars (malevolent White villain, dashing Black hero), and provided strategies to overcome bias (goal intentions to associate good with Black and bad with White) to reduce implicit preferences.

This vivid intervention is illustrated by using a TV example in Appendix 2. [Note: participants in the actual experiment read a story; they did not watch an episode of Criminal Minds.] The strategy was receiving the instruction that “the task following the story (i.e., the race IAT) was supposed to affirm the associations: White = Bad, Black = Good.”

The conclusion I draw from this impressive project is that performance on the IAT is subject to strategic control, supporting the notion that the IAT is not a pure measure of implicit attitudes. Even a brief training session is sufficient to reduce (or reverse) stereotypical preferences and associations that are supposed to be unconscious in nature (also see Hu et al., 2012; Siegel et al, 2012).



Further Reading

Open Science Framework entry on Reducing Implicit Racial Preferences: I. A Comparative Investigation of 17 Interventions

Project Implicit

"The Disturbing World of Implicit Bias..."

Human or Alien?

Human, All Too Human (AND Alien)

I'm Alive I'm Dead

Lie To Me on the Autobiographical Implicit Association Test

The Lovely Dr. ARINA K. BONES, PhD Strikes Again!


Footnote

1 Another common paradigm in cognitive psychology, semantic priming, can explain a goodly portion of the effect as well. In one study, the bias shown in IAT scores was based on statistical co-occurrence of words and concepts in the ambient culture and not on prejudiced attitudes. A discussion of those findings is beyond the scope of this post.


References

Greenwald AG, Poehlman TA, Uhlmann EL, Banaji MR. (2009). Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. J Pers Soc Psychol. 97:17-41.

Hu X, Rosenfeld JP, Bodenhausen GV. (2012). Combating automatic autobiographical associations: the effect of instruction and training in strategically concealing information in the autobiographical implicit association test. Psychol Sci. 23:1079-85.

Lai CK, Marini M, Lehr SA, Cerruti C, Shin JE, Joy-Gaba JA, Ho AK, Teachman BA, Wojcik SP, Koleva SP, Frazier RS, Heiphetz L, Chen EE, Turner RN, Haidt J, Kesebir S, Hawkins CB, Schaefer HS, Rubichi S, Sartori G, Dial CM, Sriram N, Banaji MR, & Nosek BA (2014). Reducing Implicit Racial Preferences: I. A Comparative Investigation of 17 Interventions. Journal of Experimental Psychology. General PMID: 24661055

Siegel EF, Dougherty MR, Huber DE (2012). Manipulating the role of cognitive control while taking the implicit association test. Journal of Experimental Social Psychology 48: 1057–1068.



Appendix 1

(1) Engaging with others’ perspectives
  • Training Empathic Responding
  • Perspective Taking
  • Imagining Interracial Contact 

(2) Exposure to counterstereotypical exemplars
  • Vivid Counterstereotypic Scenario
  • Practicing an IAT With Counterstereotypical Exemplars
  • Shifting Group Boundaries Through Competition
  • Shifting Group Affiliations Under Threat
  • Highlighting the Value of a Subgroup in Competition

(3) Appeals to egalitarian values
  • Priming Feelings of Nonobjectivity
  • Considering Racial Injustice
  • Instilling a Sense of Common Humanity
  • Priming an Egalitarian Mindset
  • Priming Multiculturalism

(4) Evaluative conditioning
  • Evaluative Conditioning
  • Evaluative Conditioning With the GNAT

(5) Inducing emotion
  • Inducing Moral Elevation

(6) Intentional strategies to overcome biases
  • Using Implementation Intentions
  • Faking the IAT


Appendix 2 - Vivid Counterstereotypic Scenario
...participants read an evocative story told in second-person narrative in which a White man assaults the participant and a Black man rescues the participant ( “With sadistic pleasure, he beats you again and again. First to the body, then to the head. You fight to keep your eyes open and your hands up. The last things you remember are the faint smells of alcohol and chewing tobacco and his wicked grin”).

In the therapist's office.



Escaped and running down the hallway.



Stalked by evil white sadist.



Caught and choked by evil white sadist.



FBI Profiler Derek Morgan to the rescue.



Perp in custody at the station.



The end.

5 comments:


  1. "Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies"
    http://psycnet.apa.org/psycinfo/2013-20587-001/

    ReplyDelete
  2. Rob - Thanks for the link. I didn't address the relationship between the IAT and explicit attitudes in this study. There's a huge literature on the topic, as the Oswald et al. review demonstrates. They found:

    "IATs were poor predictors of every criterion category other than brain activity, and the IATs performed no better than simple explicit measures."

    ReplyDelete
  3. The pendulum is starting to swing back, folks! For the last several years, we've been witnessing an orgy of papers, books, and TED talks on the general theme of how complex behavior by people is determined by trivial factors they know nothing about: from unconsciously smelling someone's sweat, to not washing your hands in the bathroom. One of the latest books on this theme I was asked to review was "Incognito" which, in my opinion, is a pathetic and superficial mix of psycho bubble and neurosuperficiality. The IAT was part of this trend. Now there is a push back, in part because the scientific community has gotten tired of the unreplicable BS published in social psychology journals on this theme! It is in large part untrue, and it paints a cynical and inaccurate picture of human cognition and behavior!

    ReplyDelete
  4. Well thee goes all that time and energy. What now?

    ReplyDelete
  5. I'd just like to voice agreement with Rob's comment because I think the research on which this post is based gives the IAT too much credit. I think the creators of the task (Bananji & Greenwald) were too quick to conclude that the IAT is an implicit measure of attitudes. I don't think studies that trying to reduce implicit racism (based on the IAT) deserve any recognition because the IAT has such poor criterion validity.

    ReplyDelete