One possible method is to split my reading list down the middle ahead of time, fifty-fifty, and alternate books by male and female authors. Let’s call this The Quota Method. There are a couple of obvious benefits in this approach. First, it won’t change the quality of my reading – at this moment in 2019, there are simply too many excellent books out there that alternating from one subset of authors to another won’t rule out the possibility that I can find a great read. This approach is also easily adaptable to a more expanded interpretation – I could commit to an annual reading list whose authors appear in direct proportion to that of the global demographic distribution. If my math is correct, one out of every twenty thousand books would be from an Icelandic author – what a way to spend my 270th birthday!
Of course, underlying that joke is the hidden problem of any quota – the method rules out certain great candidates by arbitrarily restricting the number of available positions. If Icelandic authors produced two great works, I would need to finish at least twenty thousand books before I could consider reading them both! This is the problem of the quota method – by taking an aggregate view of a series of decisions, it removes the possibility of getting the most value out of each individual decision. It means that instead of a reading list full of the books I decided were the best ones to read next out of all the available books, I instead chose to pick the best one out of subset A, then subset B, and so on, hoping that at the end I divided my subsets intelligently enough to produce a desired balance without sacrificing quality.
In some ways, picking a book to read next isn't so much different from hiring. In process terms, I want to get a list of candidates, learn a little bit about each, and determine the best available book to read next. And as I was in the case of my reading list, many organizations are surprised when they look at their hiring decisions in aggregate and discover a lack of diversity within their workplaces. The last part is perhaps the most intriguing part of this comparison - although I remain interested in finding ways to improve on the clumsy diversity metrics I calculated for my reading list, I am much like these organizations in how I share their reluctance to implement the only surefire way of accomplishing such an objective - a quota.
This reluctance is explainable by the incompatibility I noted above between isolated hiring decisions and the statistics used to measure those decisions in aggregate. Ask any hiring manager in the world what the goal of a hiring decision is and the answer will always be the same - to find the best person for the job. If the best candidate for a given job is equally likely to come from any non-performance based subset of the candidate pool – which is the underlying assumption of the diversity argument – then arbitrarily ruling out one subset or another will unambiguously make it less likely that the given hiring process will find the best candidate for the job. The Quota Method simply doesn't apply whenever the goal of the incremental decision is to make the best possible choice in the context of that decision alone.
But if we do not consider the aggregate metric at the time of the decision, what methods are available to help organizations reach their diversity goals? I referenced at the start that all of this thinking was a result of two recent moments, two anecdotes, and I think these anecdotes both reinforce the challenge of dealing with thorny aggregates while also pointing to the possibilities inherent in continuing to look at the problem from fresh perspectives.
The first moment was this presentation at the conference, a thoroughly researched and expertly prepared talk that covered everything I was interested in except for the one question I never hear addressed in a diversity conversation – what exactly constitutes ‘diversity’ for an organization, anyway? Let me take a crack at it. One perfectly reasonable definition is to match the demographics of the office’s home city. If this approach were applied to Boston, we would consider offices with 54% white employees as having 'average diversity' and an organization claiming itself as 'diverse' would have to match this number at the minimum. A similar alternative presented itself to me during a recent phone interview. The recruiter mentioned during our conversation that his organization regularly recruited from my alma mater, Colby, because of its strong academic record. Let's suppose we extend this comment and define diversity as matching the demographics of the average applicant's alma mater. If we use Colby as a simple proxy, we'd see that Colby’s student body is 62% white and would therefore describe an office with 62% white employees as having 'average diversity' - again, an organization claiming itself as 'diverse' would have to match this number at the minimum (1).
The roundabout point here is that if organizations don't bother defining 'diversity', they risk becoming a victim of someone else's bias. It doesn't matter whether their own process is fair or not because they won't know how to identify a biased applicant pool. It's kind of like cleaning your drinking glass to ease concerns about water contamination - it's well intended, but knowing how to identify dirty water is a much safer approach.
The unwillingness to define a diversity goal is admittedly odd to me when I consider the question from another angle - in an age when there seems to be a number for any measurable concept, why can't I find a single number out there from any company that says "this is our ideal demographic breakdown, and here's where we stack up at the moment"? I work for a company that reminds me ten different ways a day that we are data driven yet there isn't a single example I can recall where someone from my company said "this is our ideal demographic breakdown, and here's where we stack up at the moment". Why is diversity the only exception for our data driven approach?
I’m not trying to point fingers at any institution or organization. I believe everyone involved in any selection process is trying their best to overcome the various biases associated with such work. I’m certainly no better than anyone in this regard as evidenced by the breakdown of my reading list in the ‘Hello Ladies’ series. The point I’m making here is that the diversity discussion might prove more productive if the focus shifted from setting vague principles based on generally accepted truths – more diversity because diversity good – to determining the underlying inertia that prevents such a goal from being reached naturally – since we cannot accurately measure performance, we hire using secondary criteria like references, alma mater, or in-person interview performance. Like I noted above, there are some very good reasons to avoid The Quota Method as the sole criteria for making a decision. However, defining the ideal aggregate ahead of time is a good way to know when the incremental selection decisions are taking the team off the intended path and such a signal can prompt additional review of the applicant pools for any previously unseen bias.
Two years ago, I thought I made progress toward this goal when I broke down my reading list. As I noted in the series, I learned that the publishing industry tends to print more work from female authors than from their male counterparts yet male authors receive a disproportionate share of publicity. Given the important role publicity plays in how I choose the next book I read – I can only read the books I know about – I recognized the importance of finding unbiased sources from which I could freely accept reading recommendations. My most recent annual reading list didn't reflect much change in my overall results, however, so I know I need to do a little more. It's obvious that I must continue thinking about the bias built into my sources but I must also identify ways to adjust my reading selections whenever I'm waist-deep in a biased pool of new recommendations.
One approach I've implemented for 2019 is to no longer read books written by men if the only way I learned about it was from a bookstore display case. I made this adjustment for two reasons. First, I know that display cases are biased 'applicant pools' in my book selection process and therefore might over-publicize books by men at the expense of books of equal quality by women. Second, I've learned over the past couple of years that if a male author makes it into a display case and that book turns out to be any good, I'll eventually learn of it anyway from the many other sources of book recommendations in my life such as blogs, podcasts, or my social circles.
This brings me to the second moment I alluded to earlier – the podcast episode that discussed the consumer’s role in supporting an artist’s work. The podcast made me think about this lingering idea from the ‘Hello Ladies’ project because of its implication that what we support today makes the foundation for what lives on tomorrow. It's for this reason that making adjustments to how I pick the next book I read is important in a broader context. Last year, I signaled to the library that sixty percent of the books I read are written by men. How does that influence the way the library makes marginal decisions about what books to bring into its collection for next year? If my choices matter at all, there's only one plausible answer.
Anytime sixty percent of the books I borrow are written by men, I send a signal that in the future sixty percent of the books I borrow will be written by men. The only way to change such a signal is to change the way I send such a signal. And the only way I'll ever know if I'm sending the intended signal is to study what I've done in the past and make adjustments to ensure different results in the future. This strikes me as the right approach for resolving any multilayered bias problem but it's impossible to make progress without clearly defining the right goal. A new future is built just like anything else – the key step is always to set the right foundation, but there's no way to do this without first defining the vision.
Footnotes / endnotes / another needless potshot?
0. Have I ever referenced this book before?
I originally wanted to work this idea in from Michael Lewis's Moneyball:
The inability to envision a certain kind of person doing a certain kind of thing because you've never seen someone who looks like him do it before is not just a vice. It's a luxury. What begins as a failure of the imagination ends as a market inefficiency: when you rule out an entire class of people from doing a job simply by their appearance, you are less likely to find the best person for the job.1. Originally, this post was less serious, and I had more quips like this...
My hypothetical approaches to measuring diversity have their own set of problems. The city demographic idea is cute, I think, but we all know Boston isn't exactly new to allegations of racism. Should any company settle for merely matching the city's demographics? Can a diverse company emerge if it hires exclusively from such a city?
And if we agree that the best possible job candidates tend to graduate from our most prestigious institutions, then how should we consider Harvard's position as the defendant in a recent lawsuit alleging discrimination against Asian applicants in its admissions process? Should any company settle for merely matching the demographics of institutions coming under such scrutiny? Can a diverse company emerge if it hires exclusively from pools that discriminate at the point of admission?
