We Should Stop Pretending that Tech Wants to Solve its Diversity Problem
We’ve all heard the typical pitch for why we need more diversity in the tech labor force: “Look to your left. Look to your right. Before the year 2050, 2.3 of the 3 of you will need to be software engineers!”
It’s often accompanied by insightful graphics, like pie charts reminding you that women are roughly half of the population:
I’ve been hearing this same pitch for the last 15 years now. When I was a student, it would go like this: I’d be invited to attend a corporate/university-sponsored event for “Women in Computer Science,” “Women in Technology,” “Women in Mathematics,” or “Underrepresented groups in STEM,” etc. I’d go because it was a free trip, and hey, it was weirdly comforting to wait in line for the women’s bathroom for a change. I’d sit in the audience and listen to a middle-aged woman in a suit explain the “pipeline problem.” But don’t worry! Multi-million dollar corporation X is donating laptops to girls’ highschool coding clubs, so in another 15 years, the problem will be solved! And of course corporation X intends to solve the diversity problem. It’s in their best interest. How else could they expect to fill all of their open positions in 2050?
Now as an established “Woman in Computer Science/Technology/Mathematics/Cybersecurity/Finance” etc. I get invited to speak at these kind of events. And I go, and I sit in the audience and sometimes I listen to speakers before me explain the “pipeline problem.” But don’t worry! Multi-billion dollar corporation Y is donating ipads to underprivileged middle school coding clubs, so in another 20 years, the problem will be solved!
I used to think the pipeline story made sense, but for whatever reason, the diversity efforts designed to address the leaky pipeline just hadn’t really worked yet. Now I think the pipeline story does not make sense, but it does achieve exactly what it was designed to do.
The “pipeline” story goes like this: at every stage of scientific development, from middle school, to high school, to college, to grad school, to early career, to later career, etc., women and underrepresented minorities tend to drop off the STEM career path disproportionately. If we could plug those “leaks” and keep them on track, we’d have more STEM professionals in total, and also a STEM labor force with more representative demographics.
But the pipeline metaphor is misleading. It’s not wrong of course that women and underrepresented groups drop out disproportionately at several stages, and that there are many things we can and should do to address this. But a physical pipe carries more volume through to the end when its leaks are patched — the metaphorical STEM “pipeline” does not have this property.
In fact, our educational resources at several stages of the STEM progression are already under critical strain. College courses in CS are over-flowing on many campuses, and convincing more women they should sign up is not going to magically create more chairs in the classrooms. If our primary motivation is to simply produce a higher volume of trained engineers, than diversity is not the current bottleneck.
Which is why it’s not surprising to see the same corporate sponsors of diversity initiatives also participating in parallel efforts to increase training capacity. They are willing to hire out of coding bootcamps, for instance, that rather quickly retrain adults who want to pursue software engineering as a second (or third, etc) career. One could see this as evidence that they want to solve the worker shortage problem in any and all possible ways, which is complementary to the diversity pipeline story. But wait … if a significantly sized population of adults can be somewhat quickly retrained to be competent software engineers, why exactly do we need to fix the diversity pipeline at the middle school stage before we expect to fix it at the professional stage? Why are all the “diversity” branded efforts focusing on early educational stages, when real progress in addressing the skills gap is happening at later stages, bypassing the chokepoint of universities who remain resistant to respond dynamically to stark changes in demand?
This is where, if we examine closely, we can start to see the real motivations of corporations start to diverge from their pro-diversity marketing materials. They want to address worker shortages, sure. But the timelines they are most focused on are actually pretty short term. And behold — there are at least partial solutions to qualified labor shortages that work acceptably well on a much shorter timeline than 20 years! Great! Of course corporations embrace these. So why don’t they bring their diversity efforts into these shorter timelines? Well, because that’s where it would conflict with one of their other goals: not having to actually change anything about how they hire, treat their employees, design their products, and run their businesses. Incentive alignment evaporates. Giving ipads to middleschoolers doesn’t threaten the status quo in the short term. Actually redesigning a corporate structure to sincerely recruit, support, and retain a more diverse engineering workforce does.
This is not to say that currently non-diverse corporations and similarly non-diverse university engineering faculties are “anti-diversity.” They probably would prefer to magically have a more diverse workforce that functions identically to their current workforce. But established institutions just aren’t willing to jeopardize short term comfort and status to attain diversity. They are too afraid that it might require (or cause!) painful changes. There is an incredibly strong inertia at work in institutions that are doing well. Why mess with a good thing? You can see the implicit goal of excusable stasis almost nakedly in the metrics that institutions put forward to grade themselves on diversity: “We accept female applicants in the same proportion that they apply.” “The percentage of female faculty in our department is in line with our peer institutions.”
Let’s examine each of these metrics with a little more scrutiny. The first excuse is a common one that fits nicely with the pipeline story: “whatever damage to diversity has happened at the previous stages is not our responsibility to fix. So hey, we’re not making it worse, right?” But applications are influenced by an institution’s actions and culture, so this metric has some amount of self-perpetuation baked in. If an institution or entire field has a reputation that is discouraging certain groups from applying, this excuse perpetuates a viscous cycle.
Actually grappling with this is hard! How do we define non-circular metrics and processes to improve on them that are not also steeped in the biases of our history and preconceptions? Shrug. It’s much easier to say something like “we accept members of under-represented group X in the same proportions that they apply,” allow everyone to nod approvingly, and no one has to think about it too hard or be the bad guy.
The second excuse, the “we are doing no worse than our peers” excuse, is meant to suggest that the institution must be trying its best to address diversity within the incentives of a competitive marketplace. If it tried harder, well clearly that would be bad for business, otherwise someone else would have tried it already. But marketplaces aren’t perfect. They can’t simultaneously incentivize optimal behavior at every time scale, because optimal behavior clearly differs at different time scales. And peers/competitors in a marketplace tend to optimize to similar metrics and similar time scales. And hence this excuse can be used to justify arbitrarily bad effects for representation if the time horizon being considered is too short-term to truly incentivize actual progress in diversity.
My colleagues often ask me why I still go to these “Women/Unrepresented Groups in … “ events. I always come home and complain about how nothing has changed in the last 15 years. Because no one with the power to do anything about it actually wants it to change. At least, not as much as they want next quarter or next year’s metrics to look good. But I still go because 15 years ago no one stood up and told me as a student what I get up and say now — and it’s probably not what the corporate sponsors were planning. (Oh don’t worry, they don’t actually come to these things!) My case for diversity in tech goes more like this:
“Look to your left. Look to your right. Well before the year 2050, one of the three of you will be a victim of domestic violence. Now look more widely around the room. Several of you will be stalked. Technology will probably assist your stalker more than it assists you. Many of you will be harassed and bullied online. Some of you will have your social media passwords demanded by border patrol agents. Some of you will raise children whose screen addictions will have poorly understood effects on their developing brains. But don’t worry! None of this will be because you are pursuing a career in technology. Actually, it’s why we sorely need you to pursue a career in technology. And we need you now.
The tech worker shortage and pipeline story pretends that any well-trained engineer is as good as any other. But there is something the technology industry needs more sorely than your warm body at a desk chair in the future and your impressive technical mind (which you do have, by the way!) — it needs your life experience. We need there to be domestic abuse survivors and former refugees in the room of Apple engineers who are designing facial recognition as a default unlocking feature. Someone should ask — “how do we mitigate the potential for abusers and border agents to use this to force people into unlocking their phones?” We need there to be veterans in the room of Strava engineers who are deciding it’s safe to release “aggregated” location data that then reveals the confidential locations of military bases abroad. We need parents in the room of Facebook engineers who are optimizing a feature solely to “engagement.” And we need you there yesterday.
Don’t get me wrong — this is not about being contrarian. It’s not about stopping the progress of technology. It’s about making more informed decisions about risks and tradeoffs, and proactively designing features and products to better serve the needs of more communities. We could build ways for phones to exhibit multiple convincing profiles, so someone could unlock one under duress and still protect sensitive information. We could build better tools for detecting and defeating stalking apps. We could find better ways for technology to engage and facilitate healthy communication in more communities, but we shouldn’t expect to do it without the true participation of these communities.
But first, we need to actually make this the goal. I’m speaking to you today at an event that is probably designed just to check a box — so that our sponsors can say they are doing something about diversity in tech. But I’m here because I want to talk to you. Because I want to meet you. Because I want you to know that I see you, and I want you to see me. And I need to help make the next 15 years of your lives different than the last 15 years of mine. And that won’t happen if I don’t tell you the truth. It won’t happen if those of us who came before you keep pretending that it’s getting better, and that raising awareness and shaking off impostor syndrome is enough. It won’t happen if we keep accepting ipads for kids as a bribe for our silence, and letting institutions off the hook for being “in line” with their peers. And it won’t happen if I assume I know what’s best for you, without actually listening to you. So I’m glad to be here. I’m glad to have this opportunity to talk to you about my ideas and experiences and to hear about yours. And so I’d like to thank our generous corporate and university sponsors…”