A call to scientific arms on diversity
There are three things I observe many organizations getting wrong when it comes to diversity: focusing on symptoms rather than causes, looking for a one-size-fits-all solution, and being insufficiently willing to make real changes.
Today I’m going to talk about the first two. The third is important, but not relevant until we’ve figured out the others. In fact, not doing the first two may mean you spend a lot of energy trying to change things without actually thinking through whether they’re the most effective focal points for you.
For example, I am often brought in to companies to discuss topics like reducing unconscious bias, being better allies, or making more objective hiring decisions. These are great goals, but setting them assumes two things: that they are most urgent areas on which to focus when it comes to diversity in general for that specific organization. But in my experience, we have some ideas about the universals but little idea of how to prioritize them for maximum impact, and we have almost no ideas about the specifics.
Universal issues around diversity are things that are not specific to a particular organization. One universal is that all humans are biased. For example, we all tend to be more likely to interact with people similar to ourselves. The specifics refer to the fact that every organization has its own culture and instutitions that can give rise to, enhance, or subvert these biases. Because we have almost no idea how specific contexts affect the universals, we tend to take observations of the universals (humans are biased!) and try to solve them locally (unconscious bias training!) without really thinking about the environment in which it’s all playing out.
In short, we’re leaping to answers before asking careful questions. I believe we are still largely skipping over the fundamental question of: What’s causing the lack of diversity in my organization?
I encourage more companies to approach diversity using the scientific method. This means beginning with questions, answering them rigorously, being open to being wrong, and adjusting accordingly. Instead, for the most part I observe companies make an assumptions followed by unsubstantiated conclusions about what to do. Or sometimes there is no assuming, just a mindless copying of best practices.* (And, there’s rarely actual follow-up to make sure people are doing what we all think they should be doing.)
Those of you who took any middle school science class will recall the scientific method. In full disclosure, I used to think it was the most boring thing on the planet. Many years and a few brainwashing academic degrees later, I’m convinced it’s the only way to actually know what to look for and evaluate whether we’ve learned anything after we’ve looked for it. Here’s what to do.
1. Be curious about something
Most people approach diversity from a perspective of “this is a problem to solve.” It absolutely is. But it’s also very much an open question. In fact, it’s two questions: What is causing a lack of diversity (of any kind we deem important) generally in the world? And, what is causing a lack of diversity in my organization? I find that approaching diversity from the perspective of curiosity leads to more openness to answers, less defensiveness, and more objectivity. And fun! (If I keep saying “fun” will it be fun?)
2. Observation
You can’t get anywhere if you don’t have a sense of the landscape. I’m going to use the example of diversity with respect to gender because it’s an area I can speak to personally, and one on which many companies ask me to work. (Of course, there’s both a lot to do still on this front as well as many other types of diversity that need our attention). Suppose I’m interested in understanding why there aren’t more women in an organization, generally, and especially in leadership. Before I start doing anything, I need to understand what the situation really is. This step alone requires thoughtful measurement. In many large companies it can take months to get an accurate empirical picture of what the gender landscape at a company looks like (not just organization-wide, but in different roles, levels, and teams, too).
3. Theory
A theory is an idea about how facts go together. Many times when I’m asked to come in to help people figure out why there aren’t more women in their organization, they want me to tell them what concrete thing to do, like “circulate questions before meetings,” or “offer longer parental leave.” These are fine ideas (if people do them). But you’ll be much more effective if you get to the root of what’s going on and addressing that, as opposed to guessing and hoping for the best. I have my own theories, which I’ll be sharing on an ongoing basis, but people who have worked in organizations know their organization better than others, so I’d say start with their guesses, and then formalize from there.
(Yes, even the guesses of the straight, cis, white men. Especially those. Why? Because if they’re right they’ll feel good. And if they’re wrong, then you can show empirically that the assumption about why there are not many women in a company is incorrect, and you can (hopefully) put it to rest.)
4. Hypothesis
I would say 90% of the work of my Ph.D. program was learning how to generate meaningful, testable hypotheses. It’s much easier said than done and requires both creativity and discipline. It’s easy(ish) to have a casual idea about what’s happening — that’s what opinion pages in newspapers are for. It’s much harder to isolate key moving parts of that idea and transform them into statements that can be falsified using the limited and messy evidence of the real world. It gets even harder when you’re working with abstract concepts, which are at the center of most conversations about diversity.
For example, suppose you think that the reason there aren’t more women in your organization is that there’s a culture of assertiveness that’s required to get ahead, and that for the most part women aren’t socialized to behave that way, and even when they do, they’re then punished because, thanks to other socialized expectations, men will perceive them as bossy or pushy rather than assertive and confident.
This is a fine theory (depressing, but fine). But how do you test for culture, assertiveness, socialization, expectations, and gender norms? We’ll talk more about methodological tools to get at this stuff in other articles, but for now the two key things I suggest are to think about what you mean by each concept and what it would look like, and then think: If my theory is right, what else should be true? For example, if it is the case that a culture of assertiveness is keeping women from being promoted, then I would first look for indicators of assertiveness (surveys and interviews are a start, but ethnographic work like counting the number of times people interrupt each other in meetings would be useful). Then I would need to see that women and men are judged differently. Content analysis of promotion decisions and performance reviews, if they’re written, would help there. Finally, exploring where women go after they leave the firm would help rule out arguments that they’re just voluntarily exiting the workforce and it has nothing to do with this particular company. Are we having fun yet?
5. Collect data
Whoa! This is way down in the list! So many companies are eager to use data to understand what’s going on and so leap to casting about for numbers. But if you haven’t done the work to be curious, observe, and formulate theories and hypotheses, you have no idea what data to look for or start to measure. As we saw in the above example, it’s our theory about assertiveness that’s indicating the kinds of data that would be most useful.
I see this happen left and right when companies ask me what data I need to see, or what data they should collect (which can be very costly — hiring external people to build and analyze surveys for you is expensive, and I know this firsthand!). There’s effectively infinite data out there, and we could cast about forever unless we have a theory and hypothesis that points us in a certain direction. Hot tip: This is also going to help us know what kinds of analyses to conduct once we have the data.
6. Test
Actually running the statistical or experimental test is the second thing companies want to do after they leap to the data. Indeed, statistics requires a great deal of training, and having resources (that’s a cold way of saying “people”) on hand who can help conduct these thoughtfully is important. But, and I might lose my job at the NYU Center for Data Science by saying this, but I’d rather see someone do a basic analysis with thoughtfully collected data that’s high quality than a regression on a bunch of crap. More on this to come!
7. Update Theory
There’s a perception out there that science is about cold, hard facts and being correct. It’s not. Science is about asking thoughtful questions motivated by theory and then being willing to revise that theory in light of mounting evidence that suggests it might not be right. Just as we hope humans will change their behaviors to be more inclusive and all those good things, actually figuring out why there’s not as much diversity in your company on a dimension(s) you care about means being open (eager, even!) to being wrong, and adjusting accordingly.
8. Repeat
Science is also about conducting the same study and variations on that study, and a million more related studies, over and over again to clarify what we actually might be understanding, and what might be a spurious outcome of our particular study. There’s good work out there that aims to empirically and mathematically understand diversity, but much of it is in laboratories or stylized situations. I think that all organizations are possible sources of real insight on why we aren’t more diverse, and I wish we could all conduct research studies in our own organizations to slowly carve out a better understanding of what’s going on, under what conditions, and why.
And that is why I wrote this! Ok, I’m done. Goodbye!
P.S. Learning to do this crap takes forever. I don’t expect this article to instantly equip readers with the knowledge and skills to conduct scientific work, but I hope the framework of beginning with a question, imagining an answer, and then trying to prove it wrong, is useful. And, I bet you already have a bunch of people at your company trained in a lot of this stuff, anyway. (Or, call** me!)
* Of course, there are some basic fairness practices that should be in place everywhere regardless of the local ecosystem, like non-discrimination policies that are actually enforced. I’m discussing policies beyond the bare minimum. The way to think about this is — suppose you’re in an organization that has resources to spare to improve diversity. How do you know where to put those resources? I recommend scientific diagnosis rather than just casting about for best practices.
** Email or text preferably. Unscripted spoken words make me nervous.
This article is the sixth in a ten-part series on social science and diversity. Read the others here:
