 I am a behavioral scientist who studies diversity. I study diversity in many different forms, but today I'll speak mostly about racial and gender diversity with research that has been conducted almost exclusively in the United States. As a roadmap or a guide to what we'll do today, first I'll speak briefly about the workforce worldwide, the science and engineering workforce worldwide and what it looks like demographically. Then I'll speak briefly about how diversity can help drive innovation. And then I will go over three different sets of challenges regarding diversity and what the literature in my field, experimental psychology as well as organizational sociology has to teach us about diversity and inclusion. So this is the global workforce, the global PhDs. This is data collected by the National Science Foundation for 56 countries. What you'll see here is that PhDs awarded to men are in orange and to women are in green. The solid circles are PhDs in science and engineering, the ones that aren't solid are not. The y-axis here is the percentage of total PhDs granted. And along the x-axis are the countries. Now you'll see these little numbers by each of the countries. And those numbers are from the World Economic Forum's Gender Gap Index, Global Gender Gap Index. And they start at one here with Iceland and go up. They represent or measure the difference between men and women in terms of resources and opportunities in their country in health, education, economics, and politics. There are few patterns to notice here. The first one is that overwhelmingly these countries have an imbalance in terms of gender imbalance, in terms of who is getting PhDs in science and engineering. What you'll also notice here is that there are a few countries, including the United States, for which the pattern is flip-flopped, depending on whether you're looking at science and engineering PhDs or other PhDs. And that's also the case for a couple of other countries, including, I believe, Sweden and Australia. Going down the graph here, you'll also see that Ukraine, which is this dot here with the overlapping circles, actually has gender parity in PhDs in science and engineering. And if you look to the right of Ukraine, those are the very few countries which do not have the pattern where men earn more PhDs in science and engineering than women. We have very little information on race and ethnicity worldwide in terms of workforce participation in science. But we do have some data from the United States and South Africa. Looking here, what you can see at the top is that white men are overrepresented in the science and engineering workforce. White women are underrepresented in the science and engineering workforce. Other groups that are underrepresented are Hispanic males and females and black females and black males, whereas Asian females and males are overrepresented. This is particularly the case for Asian males in the U.S. You'll also find similar patterns if you look at South Africa, though, of course, the racial and ethnic categories are slightly different. Gender is, of course, the most commonly collected piece of information. So we do have data across regions in terms of who is composing the research workforce. And what you can see here is pretty dire underrepresentation of women in most of the regions. And we also have some information on socioeconomic background. These data come from the United Kingdom. And basically what you can see here is socioeconomic status as indicated by age at which parent left continuous full-time education. In other words, education of parents plotted against whether the child eventually entered the science workforce and we see a linear relationship. Finally, we have disability. Now, disability is hard to define and hard to quantify. What you can at least see from these disability numbers is that people with disabilities are underrepresented in the research, in the science workforce, at least as compared with their proportion of the population. Now, these numbers suggest some pretty drastic underutilization of certain groups. And this is perhaps a missed opportunity for many reasons, one of which is that diversity used in the right way can help power innovation. Wentz gave a talk on unconscious biases for a group of synthetic biologists. And they were very excited to tell me about a discovery that they had just made. It wasn't a synthetic biology discovery they wanted to tell me about. It was a discovery that of the prize winners in a competition, an annual global competition they had in synthetic biology, of the prize-winning teams, the ones that had women tended to win more often. And I went back and looked at the literature within experimental psychology, so literature on studies conducted more artificially in a laboratory and found that once again the studies seemed to point to groups that were composed, had some gender diversity, at least some gender diversity, tended to perform better on a variety of measures. I then looked at literature in business management. And as it turns out, one study conducted by some researchers who looked at companies in the Standard and Poor's Composite 1500 list. They looked at the representation of women in those companies in top management and found that gender diversity in top management predicted, or as they put it, resulted in an increase of $42 million in firm value. Another similar study was conducted looking at 177 U.S. national banks, looking at not gender diversity, but racial diversity, and found once again that those companies that had racial diversity, those banks that had racial diversity, tended to have enhanced firm performance. Now, for both of those studies, the companies that tended to show the biggest effects were those that were focused on innovation. As it turns out, in science, we see a hint of a similar pattern. Two economists, Freeman and Huang, recently looked at 1.5 million scientific papers that were published. They looked at the ethnic diversity of the authors and found that those papers with greater ethnic diversity were cited more and had greater impact than those with homogenous authorship, controlling for a variety of relevant factors. Of course, all these studies are basically correlational, but when we look in laboratory studies that manipulate the composition of groups, what we see is once again the benefit of diversity, such that groups that are not homogenous, groups where there are people who are different from you, cause you to actually think more and think more complexly to work harder to repair better, and those end up resulting in better team performance. It turns out that people actually also individually benefit from diversity For example, research suggests that people who live in different countries, so expats, end up actually becoming more creative according to standard measures of creativity. But diversity is difficult. Diversity is not always easy. As it turns out, what the literature also suggests is that in order to reap the benefits of diversity, you have to make sure that you provide it with a supportive context. However, three main assumptions can get in the way of that, and I'm going to go over those three main assumptions and some research that helps us to undisentangle what is going on and how we can better support diversity and inclusion. The first assumption, at least in the United States, is that in order to do diversity, you actually don't have to think very much or very hard about difference, that you actually don't need to pay too much attention to difference. And in fact, in the United States, the mantra is more that you should avoid difference or ignore difference in order to produce a fair and inclusive environment. The second assumption is that everybody experiences a class setting or a work setting in basically the same way. If you're a teacher, the students in your, I'm a professor, the students in my class, the assumption would go, are experiencing my classroom in basically the same way. So we're going to look at that assumption. The third assumption is that when there are difficulties, when there are problems related to diversity, that they are too systemic, they're too big, they're too entrenched, they're too systemic, we can't do anything about them, or rather that they're simply due to a few biased individuals. And so we actually don't need to worry too much about it. Let's tackle the first assumption, what we do about difference. A few years ago, Sodexo, the company Sodexo, was sued for racial discrimination. And in an interesting interview that was published with the CEO, Michelle Sandel, he said something that really struck me. He said that when they learned that people were unhappy and felt that they had been discriminated against, that they were surprised. So how could this have happened? How could they have been so surprised about people feeling unhappy and discriminated against such that they were taken by surprise by this lawsuit? Well, some of the research in my field can help shed some light on what possibly could be going on in such an organization. I'll give you an example of one study that I conducted. My colleagues and I went to a health care organization, a very large health care organization in the United States. And the health care organization consisted of scientists and doctors and nurses and other health professionals. We asked them about their attitudes towards diversity. In particular, we asked them about whether it was important to ignore or avoid differences. And we asked them whether instead it was important to acknowledge and embrace differences. We also asked them about their psychological engagement with their work, their standard measures for this in organizational behavior. And we did this in part because we know from other research that psychological engagement predicts productivity, turnover, and absenteeism. So those were the three measures. We had something we call in the United States color blindness or ignoring racial differences and ethnic differences. Embracing diversity or multiculturalism might be another word for that. And engagement. And we looked at 17 departments. The company was organized in 17 departments. And we correlated the majority groups' attitudes towards diversity, in this case white Americans, with the minority groups, in this case employees of color, particularly African American, Latino, and even in this case Asian employees, psychological engagement with their work and commitment to the organization. And this is what we found. In departments, if you look at the top graph, in departments where whites believed more in color blindness, minority engagement was lower than in departments where there was less color blindness. In contrast, in departments where white employees believed more in embracing and acknowledging difference, minorities were more psychologically engaged. This held controlling for a few things, including the proportion of minorities in the department. Moreover, we found that minorities in the colorblind departments tended to sense more bias. And departments in the more multicultural departments tended to sense less bias. And statistically this helped to account for this relationship. So studies have also examined what might be going on in people's minds. And as it turns out, unconscious biases and conscious biases may very well be playing a role in the results that I just showed you from the healthcare organization. In one study conducted at Dartmouth University, white college students were brought into the laboratory and not to a field with grass. And they were asked to read an article. And the article was either one that promoted ignoring differences in order to achieve racial harmony, or one that promoted embracing differences in order to achieve racial harmony. Both of them were positive. Both of them were about achieving racial harmony. Then they were asked to take what's called an implicit association test or an IAT. For those of you who aren't familiar with this test, have never seen it before, I'm going to show you an example. The job of the participant in this test is to categorize whatever shows up in the middle of the screen. In this case it would be ethnically suggestive names. So Jamal would be a name in the United States that would be a black name. Or Josh would be a white name. And they also saw positively or negatively valenced words, which I'll show you in a minute, like paradise or poison. And as quickly as they can, they have to push buttons on a computer keyboard and classify these words into one quarter or another. In those corners are different pairs of words and they switch the pairs. Here the pairs are black and good and white and bad. But they also switch them such that you have black with bad and good with white. And the computer measures how quickly you go at categorizing the words such that being faster at pairing black with bad and good with white indicates a pro-white bias or an anti-black bias. And so this is probably the most commonly used measure of implicit bias. The participants did this. They also completed a paper and pencil. Here's an example actually of how they would go about. So each time they see a word they classify it. They also completed a standard paper and pencil self-report explicit attitude task where they basically indicated how warmly they felt towards other groups. And what the researchers found was that people who read the more colorblind, the ignore differences article actually showed more implicit bias on that computer task and more explicit bias than those who read the article about acknowledging differences. So this is all one individual sitting at a computer. What about when you put individuals in interaction with one another? Studies have also looked at this. And they have found that the prescription to ignore difference and not just in race but they've also done these studies with facial disfiguration and sexual orientation and many other types of difference that the prescription to ignore difference can actually make interactions go awry. It can lead to bias in both nonverbal behavior. So in the United States that might be how much you're blinking, how far away you're sitting, how open versus closed your posture is. So bias in nonverbal behavior and bias in verbal behavior. And what some researchers did at Princeton University is that they examined the effects of priming or exposing white students to a colorblind versus a multicultural message on their ethnic minority conversation partners. And in this case the ethnic minority conversation partners were Asian and black students. So Asian and black students at Princeton University conversing with white students at Princeton University. And what they found was that the prescription to avoid race when whites were exposed to the prescription to avoid race, the minority conversation partners were more cognitively depleted. They were more cognitively exhausted by the conversation which has clear implications for their performance on other tasks. So in summary the prescription to avoid difference can actually lead to unintended negative consequences. This is consistent with research scores of studies that show that perceptions of diversity climate, students' perceptions of diversity climate on a college campus and students' perceptions of whether they have been discriminated against or experienced prejudice actually predict whether underrepresented students avoid and stay in STEM fields. You see the same thing when you look at the management literature. Perceptions of diversity climate predicts turnover intentions and actually it predicts turnover intentions among racial and ethnic minority and majority respondents. So one other study that can shed a little bit of light on how one might use difference to make a difference. This is a study conducted at Northwestern University. It's not on race, it's on socioeconomic status and the participants were either what we call first generation college students. In other words, they're the first in their families to go to college or they're continuing generation students. They're not the first in their families to go to college. These students attended panels during the first semester. In fact, I think it was in the first weeks of the semester at Northwestern and the panels were trained. In the experimental condition, the panelists were older students or more senior students, more experienced students who had been there for a while who were trained to give advice to the students but do so in a way that made socioeconomic difference explicit, that explicitly referred to class background and to first generation status. The control group were exposed to a panel that gave basically the same advice but did not expose difference, did not talk about difference, did not make difference, blatant. And then the researchers examined the grades of the students at the end of the semester. In fact, they had planned to let the study go on and examine their grades later on, but they didn't. They ended up publishing their findings right after their first wave of data collection because this is what they found. For the treatment group, this one hour intervention of going to this panel cut the gap between first generation and continuing generation students by 63% within just a couple of months. That one hour intervention, that one panel. There was no other difference between the students. They were randomly assigned to these groups. So those are some hints at how ignoring difference versus acknowledging difference can make a difference in work and educational settings. Now let's look at the second assumption. The idea that people experience a setting in the same way. As it turns out, we like to think that science is science and that if only people have the motivation and the preparation that they will join the club. This is a nice thought, but in reality when you look at the research, it's a bit more complicated than that. It turns out that a sense of belonging may be a key driver for both participation and performance. I'll give you an example from a study that was conducted by Stanford University researchers. They examined students that were incoming students. They were first year, first semester students. White students and African American students. Nothing else distinguished them. They were all students at the same university. And they gave them a one hour intervention or they were randomly assigned to the control group instead. In the one hour intervention, the treatment group, what they did was they read testimonials from older students about how when they had started school, they also felt like they had social difficulties. They felt concerns about not belonging, but that over time their confidence in their belonging grew and that things were going well now. And these were older students. So they read and interacted with these testimonials. In the control group, they did something completely unrelated. The researchers then tracked the grades and the self-reported health and well-being of these students over the next three years. And this is what they found. For white students, so those are the ones in blue, it didn't really matter, actually didn't matter at all whether they had been in the treatment group or the control group. For African American students, it mattered a lot. In fact, the treatment was so powerful that it cut the original racial gap between the students at the beginning of the study by 52%. But the effects were not just seen in GPA. They were also seen in self-reported health and well-being. So as you can see here, looking at the first graph and the third graph, which are self-assessed general health and subjective happiness, if you look at the dark orange bars, what you can see is that these African American students who were in the treatment condition reached parity with the white students in terms of their self-reported health and happiness. If you look at the middle graph, what you'll see for the African American students who were in the treatment condition, they actually reported fewer doctor's visits than everybody else in the study. So that's one study examining social belonging. And social belonging may also help to explain, not explain completely, but help to explain why historically black colleges and universities in the United States are actually stronger, in fact, the strongest producers of black STEM graduates. And it points to the importance of predominantly majority environments or predominantly white environments or predominantly non-black environments to work harder in terms of creating a supportive context. Another issue that Warren's mentioned here is the stereotype of science and how it can send signals that repels certain groups of people. So this here is an image of a white male scientist. It's actually not terribly different from what children draw when they're asked to imagine what a scientist looks like, at least in the United States. And so in the literature, we see a few different stereotypes of scientists. Number one is male. Usually white, though that is changing. We also see a stereotype depending on the field of geeky. Geeky means socially awkward, singularly obsessed with whatever their science is or whatever they're doing. Doesn't do much else with their life, which leads to another stereotype, which is that science is not collaborative. That science is something that people only do alone, which, as we know, is actually not quite accurate. Though, of course, scientists can differ a lot in terms of how much they collaborate. And finally, that scientists don't always do something that's relevant to the world, that they don't always do something that is important to solve social problems, which we also know is untrue. So my colleagues and I decided to tackle one of these stereotypes. We wanted to know what the effect was of the stereotype of the geeky computer scientist on women's participation in computer science, their interest in pursuing computer science. So we decorated a room in the Computer Science Building, the Bill Gates Computer Science Building at Stanford University, to be either conforming with the stereotype, the geeky stereotype of computer science, or to conform with a more neutral stereotype. And so here you can see when the room was decorated with the geeky stereotype in mind, there's a Star Trek poster, sci-fi books, and a pyramid of Coke cans. In the more neutral room, the non-stereotype room, we had a nature poster, neutral books, and water bottles. And we simply asked people about their interest, women and men, about their interest in pursuing computer science. And what we found was that women were significantly affected by the manipulation, such that those who answered the question in the neutral room were much more interested in pursuing computer science. We've replicated this effect not just with pursuing science in college, but also with pursuing it in jobs, jobs, teams, and companies. And we've run at least 16 studies looking at this. It's a very robust effect, and social belonging seems to be the key to it. Basically, women who are in the stereotype condition are sensing signals from the environment, even absent people. There are no people in the environment. They're sensing signals from the environment that are creating a decreased lack of interest in participating in computer science. We call it ambient belonging. Sometimes they also refer to it as talking walls. We shouldn't forget that the walls and the structure of the environment can help to create behavioral change and behavioral decisions. Another study that was done, I'll go over this very briefly, looked at the stereotype of science as non-collaborative and its effects on women. And what they found was just the simple act of having women read about a day in the life of a scientist that was full of collaboration and interactions with other people, as opposed to one that wasn't, was enough to increase women's desire to participate in science. One more note about belonging. A number of groups have sprung up. A number of nonprofits have sprung up in the United States to teach coding to underrepresented youth and girls. And many of these are actually supported by industry, including Google, has taken an interest in these. And the importance of these lies in a few things. It's not just that they're providing valuable educational and career opportunities and development to these youth, but also that while they're doing that, they're bolstering belonging, they're fostering collaboration, and perhaps most importantly, they're having students do things, develop applications that are relevant to their own lives and to their communities, which could in turn create, foster, drive innovation. Okay, so now we come to our third assumption. Is our problems of diversity so systemic that we can't do anything about them? Well, if you just listened to the first two, maybe you would come away with the idea that we should just acknowledge difference and include people, and then we have the problem solved. Researchers in organizational sociology say not so fast. There's actually a third vital component. And that third vital component is what diversity structures you put in place in your organization. So here I'm going to borrow from that literature. Researchers at Harvard studied over 800 companies. They got data from over 800 companies over 30 years. So over 30 plus years of data from these companies. And what they were able to do was see the increase in proportion of employees or managers of color. So they looked at black, Latino, and Asian employees of color, managers of color, and women of all ethnic backgrounds in these companies after an initiative had been implemented. So after a company implemented a diversity initiative, do you see an increase in the numbers? That was the goal of their study. So what did they find? First one that they talk about is having someone in charge of diversity in the company, it could be a chief diversity officer, it could be a diversity staffer. Having someone in charge who's accountable for diversity tends to increase the numbers of many of these underrepresented groups. The second one that they suggest has an important influence is having a diversity task force. Having a diversity task force is actually even a little bit more effective than just having a diversity manager. And in fact, they find that both of these contribute to the success of other diversity initiatives in the company. For example, having both of these in place can help the effectiveness of structures like employee resource groups and diversity councils that are tasked with development of employees and retention of employees. The third one, and perhaps the most important for our purposes in terms of science, is having a mentor program. And this ended up having an increase of for some of these groups almost 40% within a certain number of years. The importance of mentoring in science I think cannot be understated. If you look, for example, at science education, having a mentor often unlocks so many opportunities in science for a student. It can teach a student about what opportunities are out there. It can get them letters of recommendation. It can teach them important research skills by working in that mentor's laboratory, for example. There's the provost of Berkeley is named Claude Steele. He is a renowned African American psychologist, experimental psychologist. He recently published a book called Whistling Vivaldi. It's about stereotypes, and I highly recommend it. He wrote in his book about the importance of his mentor in terms of his own career path. And what he talked about was how his mentor appreciated him as a capable colleague, as a worthy partner, and that for his mentor, his race and class identities didn't get in the way. So that's mentoring. Interestingly, whoops, wrong button, there we go. Interestingly, these initiatives, diversity training and diversity evaluations tend not to have a big effect in terms of increasing numbers of employees of color and women in management in these organizations. Part of the reason this is interesting is because of what organizations are actually doing, which is shown here. In terms of the popularity of programs, percent of firms, which is the most popular, the one with the least effectiveness according to the research, the diversity training. That doesn't mean that diversity training is wholly ineffective, but that we need to think very carefully about the structures that we put in place. So we've talked about acknowledging difference. We've talked about inclusion. We've talked about diversity structures. There are companies and universities that have put these principles to work. So IBM, for example, is a company that noticed this, that acknowledged this over a decade ago, and they put in place a chief diversity officer and a diversity task force. And that led to the creation of diversity councils and mentoring and employee network groups. And what was important about the way that they did it was that it was fully integrated into the business. In other words, they had the support and the ear of upper management, a direct line to the CEO. But they also, through the task force and the councils, had input from local constituents at every level and across the entire organization, which led, in part, to its great success. Another company that has done similar, has implemented similar initiatives is PepsiCo. And PepsiCo has actually publicly credited these diversity efforts for its success in its billions of dollars of revenue growth, in particular with the development of new products that resulted from increased diversity. In terms of universities, I don't want to come across or leave you with the impression that the principles that I've outlined today are the only things that need to be done. This is a much more complex problem. Some universities have acknowledged this and have put together programs that take into account the principles that we've discussed today, but that also take into account much, much more, including financial aid and advising and coursework and pre-college preparation. One such example is the Meyerhoff program at the University of Maryland. And these are some examples of the 14 components that they have in place to increase the number of underrepresented minorities in science. And they've been quite successful in science, this one, yep. So, in conclusion, my aim has been to provide some principles that we can think about with research, from the literature, that we can consider in helping to increase participation that can help increase diversity and inclusion. However, it is complex. It is challenging. But what this literature suggests, all three of them, the acknowledging difference and inclusion and diversity structures, is that we have tools to do better and we can do better. And I'm happy to take your questions. Thank you.