I’ll talk about my experiences turning science into products with a social impact. Learn why a cool technology is not a business.
There is a tax and we all pay it. This tax does not build bridges. It doesn’t pave roads or pay for defense, but it is very real. In the United States, only the total combined federal tax revenue ($3.2 trillion) rivals it for scale. But it doesn’t stop at our borders. This tax is collected around the world, and every year a great deal of money and productivity is lost as a result, impeding growth and stagnating economies. And while everyone pays this tax, its most direct burden is placed on those least able to pay it, defying any concept of fairness or economic inventive. It is the tax on being different.
To understand this tax, let’s start with the story of José Zamora. Last summer, José reportedly sent over a hundred resumes a week without any response, much less an interview. So, José dropped the ’S’ out of his name and, so he claimed, everything changed. “Joe Zamora” was in demand in a way “José” could only envy.
Though José’s claims might seem exaggerated, there is ample research suggesting that his experience is real. Given identical resumes, controlled studies find that
- male names are preferred of over female,
- “Caucasian” names are preferred of “African American”,
- bias exists in simple email introductions, and
- bias persists even for matching gender and ethnicity.
Given this large and growing body of research, the classic claim “I had to work twice as hard to get where I am” is believable. Is it possible, though, to actually quantify this difference? What does it cost José to achieve the same career outcomes as Joe for the same work?
It turns out we can answer this question thanks to a data set of 122 million professional profiles collected by the HR Tech company Gild (where I was Chief Scientist). From that data, I pulled out every single José and Joe (Joe; no Josephs or any other variant). I found 151,604 Joes and 103,011 Josés. Even with just the simplest of statistics some interesting stories emerged. For example, José is more likely to become an MD than Joe, but Joe is four times as likely to reach the C-suite.
Let’s dive deeper than the surface and draw on the concept of signalling cost. A classic example of signalling cost is the peacock’s tail. Although a male peacock’s tail affords it no real survival advantage, it is a signal of the male’s fitness. The idea goes, since only the most fit males can waste energy on the biggest tails, females use it implicitly as a proxy for true fitness (much as many recruiters rely on Ivy League degrees). We can leverage the signalling cost concept here to ask, “What does it cost José to achieve the same career outcomes as Joe?”
Let’s start by pulling all of the Joes and Josés who are software developers. It turns out we still have thousands of both in our dataset, thousands of real people write code for a living. Then, let’s look at the probability that each was promoted from “software engineer” to “senior software engineer” (generic terms we used to cover a broad range of related job titles). Finally, let’s look at what sort of signals are required for José to be equally likely to be promoted as Joe. Specifically, let’s look at college degrees.
To be equally likely to get that promotion, José needs a masters degree or higher compared to Joe with no degree at all. This means that, for similar work*, José needs six additional years of education and all of the tuition and opportunity costs that education entails. This is the tax on being different, and for José that tax is $500,000-1,000,000 over his lifetime.
After diving into José’s story, I immediately applied the same analysis to other groups. Female software engineers need a Masters degree to compare equally with male’s bachelors degree. The difference is not as stark as for José, but it still means the women pay a tax of $100,000-300,000 just for a chance at the same career outcomes.
And the tax is superlinear, a black woman pays more than white woman and black man combined. Our analysis is even sensitive enough to pick up on the difference between growing up in an inclusive vs. uninclusive home for a gay professional, which is a story for some later date.
After speaking with audiences about the tax I’ve had women come up to me and say, “I’ve been considering returning to school for an MBA, but now I think I might just be paying the tax. What should I do?” Unfortunately, one person cannot refuse to pay the tax. You cannot opt out of the tax without opting out of the system. At an individual level, it is nearly as fundamental as a natural law.
The tax of being different provides a very different face on issues of bias and discrimination. The tax is largely implicit. People needn’t act maliciously for the tax to be levied. In fact, at it’s heart is (to a scientist like me) a laudable idea: prove it to me. The problem is that we are requiring different levels of proof without realizing it. “I’ll hire José…when he’s sufficiently proved his value.” Imagine how many Josés gave up long before that point in the process, disincentivized by the enormous tax they sense ahead of them.
Only at a societal or even organizational level we can say “No”. We can recognize that we all pay this tax collectively in lost potential, lost productivity, and lost lives. Companies and communities that fail to recognize and elicit the full potential of their workforce will not remain competitive. In a future post, “Engineering Environments for Success”, I will discuss how we can slowly repeal the tax on being different.
* “similar work” assessment is based on Gild’s developer ratings derived from code analysis.
“InfoCamp 2015 is jam-packed with events that will get you thinking about technology, society, and people. Throughout the day, you’ll experience…”
“Keynotes by entrepreneur Vivienne Ming on unrecognized human potential and product designer Francine Lee on the importance of personal process.Panels on data the data for good movement – with Bayes Impact and DataKind – and user-centered product development featuring designers from IDEO and SAP Apphaus.”
Being chair of the Deep Learning Summit 2015 in San Francisco was exhausting. Who knew telling dumb jokes staegically designed to bridge two unrelated talks could be so much work. From my experience, MCing and moderating are much more work and stress than speaking (which I also did). You’re responsible for making all of the speakers look good, not just yourself…which is hard enough.
The upside of Chairing the event was that I was forced to pay attention the whole time — I’ve gotten very bad about that at other conferences — giving me the chance to see some amazing work. As a scientist I loved all of the research talks but I’ll highlight a few of the demos:
- MetaMind’s fast, trainable deep learning platform was something I intend to leverage.
- On-chip deep neural networks from TerraDeep bring ubiquitous sensors into the mix.
- Can DNNs “understand” beauty and creativity? Researchers from Flickr and Yahoo showed their results in this area.
I had an amazing time talking at Bloomberg’s West Coast headquarters about the good and bad of the tech industries relationship with LGBT workers and entrepreneurs. The panel included Sarah Kate Ellis, chief executive officer of GLAAD, Kiva Wilson, diversity manager at Facebook Inc.,Patrick Chung, founding partner of Xfund, and Dominique DeGuzman, software engineer at Twilio Inc. Bloomberg’s Peter Burrows was a wonderful moderator the panel.
I was profiled in by Affinity magazine in Inspired by Brilliance.
We are excited to share the brilliance of Dr. Vivienne Ming who looks at the cost of unrealized human potential with her algorithm for redirecting recruiter hiring practices and her development of a predictive model of diabetes to better manage blood glucose levels using Google Glass.
Silicon Valley’s Embrace of the Gay and Lesbian Community – New York Times
10 Women to Watch in Tech – Inc.