This is the second in a series of three or four posts.
The first one, diverse teams perform better, explored some of the research on the measurable performance advantages that diverse teams have over monocultures. This and future posts will share real world examples about measuring and correcting for the bias that leads to a lack of diversity. I hope to build a toolkit of useful techniques that can reduce bias in recruiting, hiring, promotion, and retention.
I believe that my professional communities – high performance computing, biotech, data science, genomics, and the adjacent specialties – have a bias problem. Eventually, I hope to make that argument in some detail. Posts like this are the necessary groundwork. I hope you will bear with me.
Hang In There
Unfortunately, a lot of readers are about to check out on me.
Even though this post is absolutely not about subjective personal labels like “racist” or “sexist,” conversations about the specifics of measuring bias seem to land for some folks as accusations of bigotry. I’ve had more than a few friends and colleagues turn belligerent and hostile at this point in the conversation. I’ve been told that even wanting to talk about diversity beyond a high and fluffy level makes me sound like the “diversity police.”
Apparently nobody likes the diversity police.
Bluntly, if a person can’t bear to even talk about how one might measure and compensate for bias, it’s a pretty safe bet that their organization is rife with it.
Still, I’m not making personal accusations or calling anybody nasty names. Name calling is ineffective and misguided. It’s a waste of time.
Anne-Marie Slaughter said, “systemic bias does not require a conspiracy of men.” The Harvard Business Review built on this in their 2016 piece Designing a bias-free Organization: “…Rather than run more workshops or try to eradicate the biases that cause discrimination, companies need to redesign their processes to prevent biased choices in the first place.” Effective managers and leaders should focus on compensating for bias, rather than “naming and shaming.”
Without the power to cut funding, terminate employment, or otherwise impose consequences – naming and shaming just enrages powerful, biased people. That’s never a good scene.
Still, I acknowledge that we’re on uncomfortable ground. In a good-faith effort to keep my HPC, biotech, and genomics friends engaged, I will start off with an absolutely non-tech-centric example in which I cannot possibly be talking about them.
Let’s go to the symphony.
It sounds strange to my modern ears, but symphony orchestras used to be predominantly male. In the 1970’s, women accounted for only 6% of the players in top tier ensembles. By the 1990’s that number had risen to 21%. One major factor in that increase was the practice of “blind” auditions, where the player is hidden from the view of the judges behind a curtain or a screen.
According to the 2000 study Orchestrating Impartiality: The Impact of “Blind” Auditions on Female Musicians: “Using a screen to conceal candidates from the jury during preliminary auditions increased the likelihood that a female musician would advance to the next round by 11 percentage points. During the final round, “blind” auditions increased the likelihood of female musicians being selected by 30%.“
No quotas, no diversity police, no name calling. Just a curtain.
Let’s look at how that worked in practice:
In 1980, Abbie Conant auditioned for principal trombone of the Munich Philharmonic. The orchestra did not usually do blind auditions, but in this case one of the applicants was a relative of the decision maker, so they decided to guard against any appearance of favoritism.
By all accounts, Conant blew the doors off that audition, performing well enough that one judge leapt to his feet, saying “there’s our trombone.” On finding out who “their trombone” was, the director was nonplussed. He demoted her to second chair, paid her less than her peers, and spent years trying to get her fired. Conant spent more than a decade waging a rather epic lawsuit against the orchestra while at the same time building an international reputation as a soloist and teacher.
The Berlin Philharmonic didn’t do another blind audition for nearly 20 years.
I love the simplicity of the study above. They literally just counted the number of men vs. women who got hired before and after a process change. As the HBR article says, “Marketers have been running A/B tests for a long time, measuring what works and what doesn’t. HR departments should be doing the same.“
Of course, when we start trying to count, we find out that some job applicants are constructing their own blind-audition screens. People replace their names with initials and otherwise remove the signals that biased organizations use as unconscious cues.
What can I say? People are smart.
I assume that everybody has heard about Jo Handelsman’s straightforward and awesome study “Science faculty’s subtle gender biases favor male students.” The researchers asked faculty members to evaluate resumes for a notional lab manager that they might hire. The resumes were identical except for the gender of the applicants name.
“Faculty participants rated the male applicant as significantly more competent and hireable than the (identical) female applicant. These participants also selected a higher starting salary and offered more career mentoring to the male applicant.”
One really important result from this study is that the gender of the faculty member didn’t affect the result. “Female and male faculty were equally likely to exhibit bias against the female student.”
So spare me the line about how some particular decision can’t be biased because a woman or a person of color participated in it. That’s not what the data says.
Let’s get real for a minute: Very, very few people in any of the industries where I make a living have adopted anything even as basic as the blind audition.
We’re not even looking. We’re not counting. We’re not doing the basic work, and yet we make excuse after excuse.
We’re damn sure not doing the rigorous statistics that would be required to back up arguments about pipelines, cost-benefits of delay vs. performance, and so on. Future posts will get specific about what how we might do that.
For the moment, just rest with this question:
If your company had a “principal” position – trombone or otherwise – and the child of a powerful board member was applying for the job, what would you do to guard against the appearance of bias?
Now do it anyway.
Just don’t be surprised when the person who blows the doors off your interview process isn’t exactly who you were expecting.
2 thoughts on “Correcting for Bias”
Very cool, going to propose blinding resumes as a quick start.
It didn’t make the cut in this blog post – but another important cue for bias is the names of the institutions from which the applicant graduated. Removing university names from the early triage process does wonders for breaking our little MIT, Harvard, BU, Northeastern intellectual logjam around here.