To Become More Diverse, CS Programs Must Grow
Recently I’ve been considering two important challenges to university computer science education. The first is how to deal with rising enrollments. The second is how to improve diversity within CS. As a result of thinking about both of these goals, I’ve realized that they are linked in a way that has important consequences—particularly for efforts to improve diversity. Put simply, improving diversity in computer science requires universities both grow their undergraduate programs and disproportionately attract underrepresented students. And the fact that improving diversity is coupled to growth can make things more challenging for many departments that are struggling to grow—including mine.
Seeing the connection between diversity and growth is easy. Every year, a new group of students arrive on campus. The ones that already want to study computer science are overwhelmingly white or Asian and male for all of the well-known reasons 1. For the sake of clarity I’ll focus on gender diversity from this point forward. While computer science has a problem with most forms of diversity—including gender, racial, and socioeconomic—they all share the same link between growth and improved diversity.
Of course, one of the solutions to the diversity problem in computer science is to try and make that group of incoming freshman itself more diverse, but a potentially more powerful approach for college educators is to graduate a class that’s more diverse by improving on the unrepresentative group that we start with. So imagine that we have 100 new majors of which 80 are male and 20 are female. To get to an ideal 50/50 male to female split, we have two options:
Shrink: convince 60 men to leave computer science, or
Grow: attract 60 new women into computer science.
I have no interest in turning away anyone who wants to study computer science, so to me shrinking isn’t a viable option. That leaves us with growth. That has interesting corollaries:
It requires increasing the size of the student population substantially—by at least 60% in this example; and
it requires attracting students that are disproportionately female.
In fact, the less disproportionately female the new students are the more of them you need. In our example above the fastest way to 50/50 is with 60 new students… as long as all of them are women. If only 83% of them are female you need 120. And the gender ratio that you can achieve is bounded by the gender ratio of the new students you attract. If that group is only 30% female, then even an infinite amount of growth will only get you to 30% female.
So this ends up indicating both what we have to do (grow) and how we have to do it. Both aspects are worth more discussion. I’m going to focus on growth (what) here and largely leave the growth (how) question for another post, particularly given that it intersects with how we teach computer science at the introductory level.
So we’re not stuck with what we start with, and we have a chance to begin a fresh conversation with students of all genders and backgrounds about why computer science is so exciting.But first, is this approach really feasible? Well, not quite, since it’s unlikely that we’d be able to completely invert the same prevailing cultural norms and stereotypes that produced our 20% female entering class and manage to attract only women. However, I do think that introductory computer science courses are a terrific opportunity to help reverse stereotypes about computer science. Universities offer degree programs in many areas that high school students are unfamiliar with, and freshman and sophomores are frequently encouraged to explore courses in new fields. At UB, the result is that 65% of students change majors at least once. So we’re not stuck with what we start with, and we have a chance to begin a fresh conversation with students of all genders and backgrounds about why computer science is so exciting.
In addition, given how powerful and practical computer abilities are, it should be possible to convince students who may be headed in other directions that a few computer science courses will be complementary to their other studies 2. Once we can get them in our classrooms, the onus is on us to get them hooked and keep them there.
And finally, computer science just happens to be the most powerful tool available to anyone who wants to solve a problem or make the world a better place. Period. That’s not a bad selling point.
So growth is the answer, but it does require a department that is ready and willing to grow. Now, from the outside you might think that this would be a natural goal of any academic department. After all, one of the reasons that many faculty have chosen an academic life is that they find what they do to be extremely exciting and enjoy sharing that excitement with others. Sadly, my experience is that there are many organization factors that get in the way of growth and efforts to attract students that should be very normal departmental goals.
I expect that a common problem is instructional resources. If I had a way to double the size of our student population overnight, it would be a serious challenge to our department. While we have been growing our faculty head count, faculty hiring is a slow process and not something you want to rush if you want high-quality colleagues. And given that we are living in the shadow of the 90s tech bust, some administrators might be reluctant to rapidly add computer science faculty in case another downturn is around the horizon. (I don’t think that this is the case, and neither do several top-tier departments that have been hiring extremely rapidly recently 3.) In addition, budgeting for instructional support—in my department, mainly funding to support graduate student teaching assistants—is always lagging a year behind the numbers in our classrooms.
And then there are the administrative and institutional hurdles. Given that these are more specific to each institution, I can only comment on my experience at UB—but that has been illuminating enough. For reasons that have never been made particularly clear to our faculty 4, we have seen steadily reduced instructional support even as our student population has grown. Each year, the number of students goes up; each each year, the teaching assistant to student ratio goes down, compromising instructional quality and the balance between teaching and research in the department.
In the environment of uncertainty created by shrinking instructional resources it’s hard to embrace the growth we need to become more diverse.This is even more curious when you consider that (a) the School of Engineering, in which my department is included, is one of the only University at Buffalo units that has experienced steady growth over the past decade, (b) the computer science department is driving that trend, and (c) many of our students are international and pay higher tuition, making our growing head count even more profitable. And yet, each year as our classes get larger the number of teaching assistants per student gets smaller. In the environment of uncertainty created by shrinking instructional resources it’s hard to embrace the growth we need to become more diverse.
Pre-College Growth Masks As During-College Growth
A common thread embedded in our department’s story is that another factor limiting departmental appetites for the growth required by diversity is growth itself! There is an undeniable upswing in interest in computer science that is driving enrollments higher at universities nationwide. But, as far as I can tell, that growth isn’t making our incoming classes much more diverse: it’s taking us from 100 students with 20 females to 200 students with (maybe) 45 females. So we still have a long way to go.
While it should be something to be celebrated, pre-college growth actually creates two problems when it interacts with the during-college growth needed to improve diversity. First, pre-college growth creates the same problems I’ve described above for departments that are struggling to grow, making them even less inclined to embrace more during-college growth. When you aren’t provided the resources you need to grow, but you’re growing anyway, the last thing that you want to do is attract more students.
Second, pre-college growth can create the illusion that the department is succeeding at attracting students to CS even when it is not. Given the increasing number of students arriving at college interested in computer science, departments could be growing even when the quality of their course offerings are declining and they’re losing larger numbers of potential students. Those departments may not want to be able to separate pre-college growth from their ability to attract students on campus, since doing so points out problems that may be difficult to address.
To summarize, the connection between diversity and growth results in the following recommendations for departments that want to become more diverse:
Increase efforts to attract students on campus. Admissions changes will help a bit and pre-college outreach efforts will help a bit but what will help a lot is attracting students after they arrive on campus. Take the chance to reintroduce college students to computer science. We’re college educators—if we don’t do this, nobody else will.
Bend over backwards to attract women and other CS minorities. Counteracting our unrepresentative input pool requires unrepresentative growth that brings us back toward a healthy balance. But we should also expect that an effective introduction to computer science and a healthy departmental environment will attract men and women equally.
To a great degree, intro courses that are good at improving diversity are just good intro courses.There are many ways to accomplish these goals. Good intro courses are a big part of the solution. While they should be specifically designed with features that drive growth and diversity, happily there’s a lot of overlap between those objectives. To a great degree, intro courses that are good at improving diversity are just good intro courses. Computer science departments should also be teaching programming as service course to huge numbers of non-majors and using those courses as a chance to attract new students. And we need to make sure that we can accept students that are late to the computer science party, with a curriculum that allows them to graduate on time and to succeed even if they only start programming as freshman or sophomores. Unfortunately, I suspect that a lot of departments allow pre-college preparation to mask inadequacies in their intro courses, which not only doesn’t attract new students but actively discourages those that didn’t start programming in middle school (possibly women) or didn’t have access to a computer at home (possibly poorer students).
Messaging and on-campus outreach are also important. Departments need to send strong positive signals to women and other minorities through promotional materials, hiring, and when recognizing student achievement. Promoting student organizations that specifically try to address diversity in STEM (like Scientista) is important, and it’s equally ensure that more general computer science student groups that don’t have improving diversity as their charter (like the ACM student chapters) recognize the importance of this issue and work with the department to address it. And we need to make sure that underrepresented groups find a welcoming home in our field, which requires confronting the overrepresented groups (men) with the problems that are sometimes created by their behavior.
Growth and diversity: an approach that works
Finally, while at some level the connection is pure math, it’s nice to see it reinforced by some empirical evidence. At Harvard, the enormous success of CS50 seems to be improving diversity. And at Harvey Mudd, well-publicized efforts to increase diversity seem to also be driving growth. So efforts at growth help improve diversity, while efforts at improving diversity also drive growth. Two good things you can have together in a package deal—as long as you’re able to embrace both.