1. Increasing Women in Neuroscience (IWiN) Toolkit
Implicit Bias
This presentation should take approximately 30 minutes to complete.
Module Objective: By the end of this presentation, the audience will have a greater awareness of the effect of implicit bias on evaluation in academia and an understanding of how to combat it.
Information provided in this presentation – and the supplemental materials – is an extension of the grant funded, Department Chair Training to Increase Women in Neuroscience project.
Created by the Professional Development Committee of the Society for Neuroscience

2. Overview Global Statistics on Faculty Salaries by Sex
Diversity and Scientific Excellence
Implicit Bias (Schemas): What Is It, How Is It Measured?
Evidence that Implicit Bias Affects Evaluation
Strategies for Breaking the Cycle

3. Some Global Statistics on Women in SET
Participation refers to the enrollment of women in undergraduate and graduate programs (tertiary science and engineering enrollment)

4. The infographics and data displayed on this slide were obtained from the 2014 Life Sciences Salary Survey. The complete survey can be found on Data for the Life Science Salary Survey was collected by The Scientist via a Web-based survey, which was open from March 10 to July 21, A total of 5,334 usable responses were submitted and considered as part of the survey. The 2015 data is now available at the same web address and confirm the continuing existence of the gender-based salary gap.
The data depicts jobs in the life sciences, broadly defined. Respondents were employed in technical positions, as tenure-track and non-tenure-track faculty, in pharmaceutical and biotechnology companies, and clinical research.

5. Full Time Faculty Member Salaries in the United States
The data was obtained from the 2013 AAUP Salary Survey. The complete survey can be found on the AAUP.com website.

6. From Life Sciences Salary Survey 2014; The Scientist, Nov. 2014, p. 55
The graph and data were obtained from the 2014 Life Sciences Salary Survey. The complete survey as well as previous and later (2015) reports can be found on
As stated in the 2014 Life Sciences Salary Survey report, currently “the gender gap in the U.S. is only found in high-ranking positions. While women and men at early- and mid-career stages—from graduate student to associate professor—received nearly equal pay, women professors only received 88 percent of the income male professors did. And because more men hold these high-ranking positions, the overall pay gap in the U.S. is magnified.”
For more data on the “leaky pipeline”, documenting declining participation of women and URMs in academia over the past three decades as they progress up the professional ranks from graduate students to faculty ranks, and validated strategies for improving recruitment of such individuals, see the IWiN Toolkit Module “Recruitment and Retention.”
From Life Sciences Salary Survey 2014; The Scientist, Nov. 2014, p. 55

7. The graph and data were obtained from the 2014 Life Sciences Salary Survey. The complete survey can be found on The data depicts jobs in the life sciences, broadly defined. Respondents were employed in technical positions, as tenure-track and non-tenure-track faculty, in pharmaceutical and biotechnology companies, and clinical research.

8. Gives us access to talent currently not represented.
Why do we need to recruit a diverse faculty in order to attain excellence?
Gives us access to talent currently not represented.
More perspectives are taken into account in devising solutions to problems.
Heterogeneous groups are more effective in problem solving, demonstrate greater creativity, and improve the vigor of a scholarly community (see references).
Women make up slightly more than 50% of the population but are not 50% of the workforce. Likewise, under-represented minorities do not participate in the scientific workforce at levels equivalent to their percent of the population.
References are as follows:
Ely & Thomas (2001). Administrative Quarterly 46 (2),
Page, S. (2007). The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies: Princeton University Press.
Temm (2008). In L. Schiebinger (Ed.), Gendered Innovation in Science and Engineering (pp ).

9. Why is it difficult to recruit for diversity and excellence?
It is tempting to believe that discrimination against certain groups is a thing of the past, or is only practiced by a small set of uninformed people.
Research shows that we all – regardless of the social groups we belong to – perceive and treat people differently based on their social groups (race/ethnicity, gender, sexual orientation, disability, etc.).
Valian (1998) Why So Slow? The Advancement of Women. Cambridge: MIT Press, p. 280.

10. Schemas: Non-conscious Hypotheses
Schemas (expectations or stereotypes) influence our judgments of others (regardless of our own group).
All schemas influence group members’ expectations about how they will be judged.

11. Schemas are… Widely culturally shared
Both men and women hold them about gender.
Both whites and people of color hold them about race/ethnicity.
People are often not aware of them.
Applied more under circumstances of:
Ambiguity (including lack of information)
Stress from competing tasks (e.g., job and family responsibilities)
Time pressure
Lack of critical mass (minority status)
Fiske (2002). Current Directions in Psychological Science, 11,
Steele (1997) American Psychologist, 52 (6):613‐629

12. - Slide provided by STRIDE: University of Michigan ADVANCE
What is the effect on applicants -- aspiring students and potential faculty?
How is it that people committed to diversity made such a web page? It was clearly not done intentionally, meaning that there was an unconscious element.

13. Unconscious Bias: Even on Google
“I was arguing with my girlfriend about women not inventing anything useful. In an attempt to prove me wrong, she Googled “She invented” only to have it ask, “Did you mean ‘He invented’?’” – submitted by Tom Boutcher to dig.com: May 6, 2007
Also works with “discovered,” “calculated,” “analyzed,” and even “led.” However, if you type in something like “she cried,” it does NOT ask if you meant “he cried.”
- Google manually corrected this problem in 2007, when it was brought to their attention.

14. Gender-Science on Project Implicit
Courtesy: Frederick L. Smyth, UVA, 2013
Project Implicit uses a simple response time task administered online to assess participants’ association of specific traits (e.g., aptitude for science or math) with factors such as gender or race. As described on the web site: “Project Implicit is a non-profit organization and international collaboration between researchers who are interested in implicit social cognition - thoughts and feelings outside of conscious awareness and control. The goal of the organization is to educate the public about hidden biases and to provide a ‘virtual laboratory’ for collecting data on the Internet.” Over time, responses have been collected from hundreds of thousands of individuals from more than 30 countries.
70% of respondents perform better (shorter response time) with the stereotypical pairing (male=science; female=liberal arts) while only 10% do better with the counter-stereotypical pairings.

15. Gender-Science on Project Implicit
The vast majority of respondents show a bias of science=male.
The white bars are the 20% whose scores don’t show a gender bias.

16. Same for Men and Women (unless…)
If we look at men and women separately, there is essentially no difference in the pattern of implicit stereotype strength.

17. Academic Identity Matters
The Y axis represents the degree of bias against women in science. A score of 0 would indicate no bias. Scores above 0 represent increasing degrees of bias for male=science. Data are from US undergraduate students.
Men in science majors have the strongest stereotypes of all men. Women in non-science majors have the strongest stereotypes of all women. When we compare men and women in majors at far right, the gender gap in implicit bias is among the largest psychological sex differences we see, a difference of about .8 standard deviations.
So we see that our personal identities matter. Specifically, in this case, the combination of two identities is at work: gender and academic discipline.

18. Greater 8th-grade Boys’ Advantage correlated with greater country-level implicit bias, r = .60
Data courtesy of Nosek, Smyth et al., 2009, PNAS
The graph displays the results for the 34 nations considered as part of the Trends in International Mathematics and Science Study (TIMSS) in The correlation indicates that in countries where boys out-scored girls by greater amounts, citizens of those countries taking the gender-science IAT (implicit association test) exhibited greater bias. For those who think in terms of correlation coefficients, it was .6.
In the published report Smyth et al. speculated that each factor influences the other. That is, we think conditions of gender equity on the ground, whether indicated by test performance differences or different gender proportions among science faculties, probably influence citizens’ implicit biases. And, in the other direction, implicit biases may shape students’ self-identification and engagement with different subject areas and lead to different average group performances.

19. Implicit Bias Can Affect Evaluation
Blind auditions
Evaluation of student inquiries
Evaluation of resumes
Evaluation of CVs
Evaluation of fellowship applications
Letters of recommendation

20. Blind Auditions: Gender
Records from major US symphony orchestras from : Audition data from 14,000 individuals show the use of a screen increases the probability that a woman will advance from preliminary rounds by 50%.
Roster data from 11 major orchestras show the switch to blind auditions accounts for 30% of the increase in the proportion of women among new hires.
- Reference: Goldin & Rouse (2000) The American Economic Review, 90, 4,

21. Implicit Bias: Responses to White Males vs. Other Students
Numbers in parentheses on y axis labels are the percent of inquiries from women and minorities regarding research opportunities that received a response from faculty in each discipline. The x axis (Size of the Discriminatory Gap) represents the difference in response rate for white males compared to women and minorities.
Data are from K. Milkman et al., Journal of Applied Psychology, 2015, Vol. 100, pp. 1678–1712

22. Unconscious Implicit Bias: Hiring and Promotion
Both men and women are significantly more likely to rank a perceived man higher than a perceived woman, using identical resumes.
Fidell, L. S. (1970). Amer. Psych. 25,
Steinpreis, R.E., Ritzke, D., and Anders, K.A. (1999). Sex Roles, 41,
Moss-Racusin et al., PNAS, “Science faculty’s subtle gender bias favors male students”,

23. Evaluation of Identical CVs: Gender
When evaluating identical application packages, both male and female University psychology professors preferred 2:1 to hire “Brian” over “Karen” as an assistant professor.
When evaluating a more experienced record (at the point of promotion to tenure), reservations were expressed four times more often when the name was female.
Karen
Brian
- Slide provided: STRIDE: University of Michigan ADVANCE
Science faculty’s subtle gender biases favor male students
Corinne A. Moss-Racusina,b, John F. Dovidiob, Victoria L. Brescollc, Mark J. Grahama,d, and Jo Handelsmana,
PNAS (2012)
Steinpreis, Anders, & Ritzke (1999) Sex Roles, 41, 509.

24. Evaluation of Identical Resumes: Race
Applicants with African American- sounding names had to send 15 resumes to get a callback, compared to 10 needed by applicants with white-sounding names.
White names yielded as many more callbacks as an additional eight years of experience.
Jamal
Greg
- Slide provided: STRIDE: University of Michigan ADVANCE
Bertrand & Mullainathan (2004) American Economic Review, 94 (4),

25. - Men are perceived as more worthy of mentoring.

26. - Interestingly, though a female name on the portfolio was a liability on every competence-related outcome, it was a plus for ratings of likeability.

27. Evaluation of Fellowship Applications: Gender
“…the success rate of female scientists applying for postdoctoral fellowships at the [Swedish Medical Research Council] during the 1990s has been less than half that of male applicants.”
Women applying for a post- doctoral fellowship had to be 2.5 times more productive to receive the same reviewer rating as the average male applicant.
- For U.S. based data, review the data surrounding the NIH Pioneer Awards
Similar findings:
USA/GAO report on Peer Review in Federal Agency Grant Selection (1994)
European Molecular Biology Organization Reports (2001)
NIH Pioneer Awards: Journal of Women’s Health (2005) & Nature (August 2006)
Wenneras & Wold (1997) Nature, 387, 341.

28. Differences in Letters of Recommendation for Successful Medical School Faculty Applications
Letters for women :
Shorter
More references to personal life
More “doubt raisers” (hedges, faint praise, and irrelevancies)
“It’s amazing how much she’s accomplished.”
“It appears her health is stable.”
“She is close to my wife.”
Letters for men:
Longer
More references to: CV
Publications
Patients
Colleagues
Trix & Psenka (2003) Discourse & Society, Vol 14(2):

29. Critical Mass Affects the Use of Schemas
When there are many individuals, we differentiate among them and cannot rely on group-based schemas.
In both experimental and field settings, increasing the female share of those being rated increased ratings of female applicants and employees.
References:
Valian (1998) Why So Slow? The Advancement of Women. Cambridge: MIT Press, p. 280;
Heilman (1980) Organizational Behavior and Human Performance, 26: ;
Sackett et al (1991), Journal of Applied Psychology, 76(2):

30. The Gender Composition of Elite Biological Laboratories in the U.S.
Elite laboratories include those funded by the Howard Hughes Medical Institute (HHMI), members of the National Academy of Science (NAS), and winners of prestigious career awards such as the Nobel Prize. The weighted average percentages of female trainees in laboratories with (A) male PIs and (B) female PIs who have achieved certain career milestones are displayed. Major career awards that were counted for this survey are listed in Table S2 of the published paper. *P < 0.05; **P < 0.005; ***P < (Wilcoxon rank sum test).
Sheltzer J M , and Smith J C PNAS 2014;111:

31. Feeder Laboratories Train Fewer Female Postdocs
Feeder laboratories are defined as those that successfully trained postdocs who won recent faculty job searches at top universities.
The PIs of feeder laboratories were significantly more likely to be HHMI investigators, NAS members, or have won a major research award relative to the pool of all PIs (Fig. 3A). For instance, 13% of the professors in the dataset were members of the NAS, but 58% of feeder laboratory professors were NAS members (P < , Fisher exact test).
(A) The percentages of all faculty members (black bars) and faculty members who have recently trained new assistant professors (white bars) who have achieved certain career milestones are displayed. ***P < (Fisher exact test).
Sheltzer J M , and Smith J C PNAS 2014;111:

32. Feeder Laboratories Train Fewer Female Postdocs
(B) The percentages of female postdocs employed in different laboratories are displayed. The black bar represents all PIs, the gray bar represents PIs who have recently trained at least one new assistant professor of either gender, and the white bar represents PIs who have recently trained at least one new female assistant professor. **P < .005; ***P < (Fisher exact test).
(C) The percentage of female postdocs employed in different laboratory types subdivided by the gender of the PI is displayed. ***P < (Fisher exact test).
Sheltzer J M , and Smith J C PNAS 2014;111:

33. Feeder Laboratories Train Fewer Female Postdocs
(D) The percentage differences in female postdocs employed by different PIs are displayed. The axis at x = 0 represents employing female postdocs at a rate proportional to their representation among all laboratories in this survey.
Sheltzer J M , and Smith J C PNAS 2014;111:

34. Impact of Schemas on Leadership
With single sex groups, observers identify the person at the head of the table as the leader.
With mixed sex groups
a male seated at the head of the table is identified as the leader.
a female seated at the head of the table is identified as the leader only half the time (and a male seated somewhere else is identified the other half).

35. Impact of Schemas about Parenthood
Assumptions about the implications of motherhood for women’s career commitment have consequences, despite recent data showing that:
Women academics who marry and have families publish as many articles per year as single women.
“…net sex differences in productivity are small to nil once other personal characteristics, structural settings, and facilitating resources are taken into account.” – Xie and Shauman
Yu Xie and Shauman (2003) Women in science: Career processes and outcomes.
Cole and Zuckerman (1987) Scientific American 256 (2),
Boushey (2005) Center Economic Policy & Research

36. Schemas do…
Allow efficient, if sometimes inaccurate, processing of information.
Often conflict with consciously held or “explicit” attitudes.
Change based on experience/exposure.
Nosek, Banaji, & Greenwald (2002). Group Dynamics: Theory, Research and Practice, 6,
Fiske, Cuddy, Glick, & Xu (2002). Journal of Personality and Social Psychology, 82(6),

37. Strategies for Breaking the Cycle
Increase conscious awareness of bias and how bias leads to overlooking talent
Implicit Association Test: Greenwald, McGhee & Schwarz, 1998
Develop more explicit criteria (less ambiguity)
Alter departmental policies and practices
For examples on how these strategies have been implemented with success, visit the IWiN Toolkit: Candidate Recruitment and Evaluation. Additional success stories are available within the NSF-funded IWiN interactive course modules, available here:

38. The bias of female students against women as leaders was increased in women who attended a coed college as freshmen whereas that of those attending the women’s college was reduced to zero bias.
Dasgupta found that it wasn’t the type of college, per se, that made the difference, but that the women’s college women interacted with more female faculty.

39. Take Homes Diversity fosters excellence
Implicit bias affects evaluation
Implicit biases can change
Self-concepts and environments matter
We’re off to a good start, but more longitudinal research on development and influence of implicit bias in STEM is clearly in order.

40. Acknowledgements NSF ADVANCE PROGRAM
Pamela Raymond, ADVANCE Program at the University of Michigan, Strategies and Tactics for Recruiting to Improve Diversity and Excellence (STRIDE)
Society for Neuroscience
Fred Smyth, University of Virginia