The Science of Unconscious Bias Toni Schmader Department of Psychology University of Arizona

Outline of Presentation Understanding unconscious associations Demonstration of our biases How unconscious bias affects our behavior Breaking free of biases

Being of Two Minds Reflective system for controlled processing Conscious, explicit Effortful, requires motivation Takes more time Reflexive system for automatic processing Often unconscious, implicit Requires little effort Fast Different neural structures distinguish the two Satpute & Lieberman (2006) I look around the room and see a group of people are in the business of using of the reflective system. As academics, we like to think of ourselves as reflective thinkers. The type of people who always act based on conscious consideration of the evidence. But we need the reflective system as much of the time our brain runs on autopilot. This is easily seen for everyday tasks. Like when your good intentions to stop at the market on the way home from work get foiled because your brain ran the more automatic routine of driving straight home. This reflexive system is quite functional – it allows us to accomplish things on based on the well-practiced routines or automatic associations we have learned from years of experience navigating our social worlds. But when it comes to our interactions with others, our more automatic and reflexive system can sometimes leads to biases, that our more reflective system would not commit.

The Reflexive System Uses Implicit Associations Cognitive links between concepts that co-vary Bring one to mind, others are activated Activation can happen unconsciously ...can be at odds with conscious goals …can influence attention, perception, judgment and behavior Our brains are exquisitely skilled of observing co-variation in the world. When we see time and time again that the presence of dark clouds precedes the fall of rain, we learn to associate clouds and rain. When the state legislature meets, we might learn to expect budget cues. The repeated exposure to this covariation means that over time, an implicit cognitive association is formed between these two concepts…such that when I say legislature, many of us might even wince in anticipation of the negative news we fear will follow.

LAUNDRY

The procedure is quite simple The procedure is quite simple. First, you arrange things into different groups. Of course, one pile may be sufficient, depending on how much there is to do. If you have to go somewhere else due to lack of facilities, that is the next step; otherwise you are pretty well set. It is important not to overdo things. That is, it is better to do too few things at once than too many. At first the whole procedure will seem complicated. Soon, however, it will become just another facet of life.

XXXX XXXX COW

Count the Number of Passes between White vs. Black shirted Players Neisser (1979)

Unconscious Gender Biases Unequal gender distribution of men and women in certain roles creates implicit associations Eagly (1987); Glick & Fiske (1996) With domains… Work = male; Family = female Science = male; Arts = female That generalize to traits… Male = independent, competent Female = cooperative, warm Our brains are exquisitely skilled of observing co-variation in the world. When we see time and time again that the presence of dark clouds precedes the fall of rain, we learn to associate clouds and rain. When the state legislature meets, we might learn to expect budget cues. The repeated exposure to this covariation means that over time, an implicit cognitive association is formed between these two concepts…such that when I say legislature, many of us might even wince in anticipation of the negative news we fear will follow.

One Way to Measure Unconscious Bias The Implicit Association Test (IAT) Greenwald, McGhee, & Schwartz (1998) Measures strength of association between concepts Based on premise that associated concepts will be easier to categorize together

Men and Women both Show Implicit Gender Biases Association of math = male & arts = female Nosek et al. (2002) Association of men = independent & women = communal Rudman & Glick (2001) Men and women showed equivalent tendency to unconsciously associate men with agency and women with warmth. On a more conscious measure, men gender stereotypic associations were stronger than were women’s. Greenwald et al. (2003) describe the scoring algorithm for calculating the IAT effect in detail. It involves calculating the difference in average response latency between the two sorting conditions and dividing by the standard deviation of all latencies for both sorting tasks. Thus, the IAT score (called D) is a cousin of Cohen’s d calculation of

Data on the IAT (Nosek, Banaji, & Greenwald, 2005) In comparison, effect size for gender differences in complex mathematical problem solving: d = .29 Hyde, Fennema, & Lamon, 1990 From method paper in PSPB. Data collected on the web (note, that some participants are excluded from each of these samples)

Implications for Behavior Implicit racial biases predict… Amygdala activation (fear response) Phelps et al. (2000) Lower performance ratings Amodio & Devine (2006) Avoid the other group Amodio & Devine (2006); Phills & Kawakami (2005) More negative interactions Dovidio et al., (2002); McConnell & Leibold (2001) Phelps shows that amount of IAT bias predicts amount of amygdala activation when viewing (unfamiliar) black faces. Avoidance studies show that people with strong IAT bias are faster to push a joystick away from them when presented with black faces vs. white faces; or sitting farther away from the person Amodio and Devine show the difference between attitudes (predicting preference and behavioral association) and stereotypes (predicting expectancies). Performance ratings on essay the person presumably wrote or

His view of the Interaction Her view of the Interaction Dovidio et al., 2002 His view of the Interaction Predicted What Was Said r = .36* r = .40** Degree of Explicit Bias “I’m not prejudiced” Degree of Implicit Bias “Black = Bad” Her view of the Interaction Predicted How it Was Said r = -.41** r = .34*

Implications for Behavior Implicit gender biases … Predict biased ratings of job candidates Rudman & Glick (2001) Might be manifested in letters of recommendation Schmader et al. (2008), Trix & Psenka (2003) Men are more often described with superlatives & as having ability Women are more often described as working hard Can contribute to women’s weaker association with math Even among math & science majors Nosek et al. (2002) Among men & women, independent = male bias predicts… Rating assertive male candidate as more qualified Rating assertive female candidate as lacking warmth Rudman & Glick (2001) Among female college students, math = male bias predicts… An unconscious negative attitude toward math Poorer performance on math SAT Even among women majoring in math and science Nosek et al. (2002)

A Two Strategy Solution Change Implicit Associations Consciously Override Biases Unconscious Biases Judgment & Behavior

1) Overriding Unconscious Bias Be motivated to control bias Be aware of the potential for bias Take the time to consider individual characteristics and avoid stereotyped evaluations Instead of having the frame of avoiding being prejudice, have the frame of approaching being fair.

Example When writing evaluations, avoid: 1. Using first names for women or minority faculty and titles for men (Joan was an asset to our department.” –vs.- “Dr. Smith was an asset to our department.”) 2. Gendered adjectives (“Dr. Sarah Gray is a caring, compassionate physician” –vs. – Dr. Joel Gray has been very successful with his patients”) 3. Doubt raisers or negative language (“although her publications are not numerous” or “while not the best student I have had, s/he”) 4. Potentially negative language (“S/he requires only minimal supervision” or “S/he is totally intolerant of shoddy research”) 5. Faint praise (“S/he worked hard on projects that s/he was assigned” or “S/he has never had temper tantrums”) 6. Hedges (“S/he responds well to feedback”) 7. Unnecessarily invoking a stereotype (“She is not overly emotional”; “He is very confident yet not arrogant”; or “S/he is extremely productive, especially as someone who attended inner city schools and a large state university”

A Two Strategy Solution Change Implicit Associations Consciously Override Biases Unconscious Biases Judgment & Behavior

2) Changing Unconscious Bias The effectiveness of education (Rudman et al., 2001) Prejudice Seminar was taught by black male professor, control class was a methods course taught by a white female professor. IAT measured at the beginning and end of semester. Non-Black participants show significant decrease in negative implicit associations with blacks after taking the course. Other analyses suggest change due to liking the professor and greater contact with Blacks.

2) Changing Unconscious Bias The effectiveness of education (Rudman et al., 2001) The effectiveness of exposure (Dasgupta & Asgari, 2004) Sample of college freshman women tested at start of freshman year and one year later IAT is leadership = male No difference between women at the two schools in IAT at time 1, but significant difference at time 2. Replicate this effect in an experiment manipulating exposure to female leaders

2) Changing Unconscious Bias The effectiveness of education (Rudman et al., 2001) The effectiveness of exposure (Dasgupta & Asgari, 2004) Sample of college freshman women tested at start of freshman year and one year later IAT is leadership = male No difference between women at the two schools in IAT at time 1, but significant difference at time 2. Replicate this effect in an experiment manipulating exposure to female leaders

Take-Away Points Implicit bias is distinct from conscious motivation We all have these biases due to cultural exposure They can affect behavior unless we override them They can be changed with education and exposure

Questions, comments, insights? Take other Implicit Associations Tests Online: https://implicit.harvard.edu/implicit/

Workplace Conversations 18 male and 18 female STEM faculty 88% response rate Electronically Activated Recorder (EAR) Sampled audio snippets during 3 workdays Participants complete workplace surveys of job satisfaction and disengagement Coding Conversational snippets transcribed & coded for content

-.42a .72b .44bc -.18acd -.27a -.23abd .33bc .41c -.26a .39b .51b Conversations with male colleagues Conversations with female colleagues Male Participants Female Participants Research talk… Job disengagement -.42a .72b .44bc -.18acd Job satisfaction -.27a -.23abd .33bc .41c Collaboration talk… -.26a .39b .51b .06ab -.24abc -.50ab .03abc .31ac Social talk… .51a -.50b -.22bc .50ad .29a .58ab -.25ac -.29cd

Conclusions Female faculty feel greater job disengagement and less satisfaction… to the degree that they discuss research and collaboration and do not discuss social topics …with their male colleagues