Wednesday, July 9, 2014

Your Unconscious Gender Bias Could Kill You

Today’s guest blogger is Stella Offner. She is a Hubble Postdoctoral Fellow at Yale University.
If you’ve been a long-time (or even recent reader) of AAS Women, you will be familiar with the many perils of unconscious bias (1). You will be aware that unconscious bias related to gender can result in unintended discrepancies in women’s salary, citation count, award recognition, funding, mentoring opportunities, and of course, flat-out discrimination.  All these things are bad for women generally and for equality in science, specifically. Just in case you are still not convinced that gender bias is not a big deal and doesn’t apply to you, did you also know that your unconscious gender bias could kill you? Seriously.

Now that I have your attention, a recent study published in the Proceedings of the National Academy of Science (2) examined the number of deaths that occurred as a result of hurricanes with female names versus those with male names. They found that there were two to three times as many deaths due to hurricanes with female names. Statistically speaking, female-named hurricanes are not stronger on than male-named hurricanes (the gender of the name alternates by convention). Instead, the authors claim that this deadly result is due to individual’s behavior in response to hurricane warnings for female-named hurricanes versus male hurricanes. That’s right, the study suggests that people, on average, expect female-named hurricanes to be nicer and less violent than male-named hurricanes. This could only be an unconscious reaction because hurricanes are in no way like men and women. They do not have masculine or feminine traits; female hurricanes don’t gestate and nurture little hurricanes.  Yet, there it is.
Let me take a step back. Deaths that result from natural disasters are tragic. I am not claiming, nor does this study claim, that any particular person who has died is gender biased, unconscious or otherwise.  A person’s decision in response to weather warnings involve a very complex process, so it is unclear how unconscious bias percolates into human behavior. Perhaps, the hurricane severity is reported very subtly differently by news outlets. Perhaps, the hurricanes are discussed very slightly differently in the media or among individuals. Even small differences could statistically lead more people to ignore warnings, resulting in more people at risk.
If you are an astronomer or other data scientist, the next question you may be asking yourself about this study is: how robust is this statistical conclusion? After all, the correlation between name-bias and death must have many complex factors: this is not the time to mess around with chi-squared linear fits. The authors removed the two worst statistical hurricane outliers, Katerina and Audrey, both of whom were responsible for at least an order of magnitude more deaths than the typical hurricane, male or female (including them in the analysis makes the correlation stronger). 
The authors next performed six experiments to follow up the archival study and confirm their hypothesis. They used hundreds of volunteers (college students) to investigate unconscious-bias-based reactions to hurricane warnings. One group of participants rated the severity of 10 hurricanes of which 5 had feminine names. Another group were given a weather map for either ‘Hurricane Alexandra’ or ‘Hurricane Alexander’ (identical maps) and asked to rate their severity. A third group was given identical information about either ‘Hurricane Christina’ or ‘Hurricane Christopher’ and asked how likely they would be to evacuate. All the experiments found the same result: hurricanes with female names were rated to be less severe and were less likely to lead to evacuation than hurricanes with identical strengths given male names. The good news is that when participants were directly asked to choose which of a male-female hurricane pair was more risky, the results were evenly split; there was no conscious bias.
So what can we take away from this study? The authors propose that hurricanes should no longer be given gendered names. This would save lives merely by changing a naming convention. But what do we do about the cause of the problem? Unfortunately, changing our own names to make them un-gendered is not a realistic solution. This study makes me wonder about all the ways unconscious bias affects overtly gendered-named female scientists, such as myself. Maybe we are expected to be ‘nicer’ mentors and colleagues than our male colleagues, which may lead to a backlash if we are instead similarly collegial? Indeed, two studies by Yale Professor Victoria Brescoll found that women were penalized for stereo-typically unfeminine behaviors like expressing anger or ambition, while men were rewarded (3,4). Similarly, are severe global warming studies or other scientific warnings less seriously received if individuals project their unconscious perceptions of female personality onto the scientific conclusions? Are grant proposals attached to more feminine names judged more harshly if they appear too ambitious, aggressive or immodest?
If assigning gender where none exists is alone sufficient to change individual behavior, there are likely many more examples that have not been explicitly identified. Consequently, this suggests that even those of us who are aware of unconscious bias and actively policing ourselves must exercise vigilance to its unexpected influence.
And of course, the final lesson is that if Hurricane Betty is bearing down on the coast, we should all definitely get out of her way.
(2) Jung, Shavitt, Viswanathan and Hilbe 2014,
(3) Okimoto & Brecoll, “The price of power: Power-seeking and backlash against female politicians”, 2010 Personality and Social Psychology Bulletin.
(4) Brescoll & Uhlmann, “Can an angry woman get ahead? Gender, status conferral, and workplace emotion expression”, 2008, Psychological Science.