Students in Allison McDonald’s course Data Science in Human Contexts debate an issue with real-life applications: Should Amazon allow law enforcement officials to access Ring video without the owner’s consent? This focus on how data science can be leveraged in various fields has helped BU’s Faculty of Computing & Data Sciences recruit and retain women as faculty and students.

Science & Technology

Women Have Been Shut Out of Computer and Data Science. BU Is Opening the Door

The University’s Faculty of Computing & Data Sciences has been uniquely successful in recruiting—and retaining—women in the field

Across the United States, women make up roughly half the workforce and roughly half the population. But, zoom in on the STEM fields (science, technology, engineering, and math), and it’s a starkly different story. In data science, for example, only about 15 to 20 percent of professionals in the field are women. 

It’s a trend that cuts across all levels of employment in these fields: less than 20 percent of cutting-edge start-ups are created by a founding team that includes a woman, and just 3 percent of those founding teams are composed entirely of women. The same holds true for many higher education programs. Women’s shares of computer science degrees peaked in the 1980s, around 35 percent, during the early days of Microsoft and Apple, then dropped precipitously—by 2020, it was closer to 20 percent. 

Boston University, however, is bucking this trend. The University’s Faculty of Computing & Data Sciences (CDS), which launched in 2021, has been uniquely successful in recruiting—and retaining—women in the field. Women made up 46 percent of its inaugural class, and it’s grown from there. Of the Class of 2028, 47 percent are women. 

Emma Steel (CAS’25, CDS’25), a senior in CDS, can easily recall being one of the only women in her Florida high school computer science classes. Being at BU, she says, “is refreshing—and really exciting.” 

Azer Bestavros, associate provost for computing and data sciences, frames it as a societal benefit: “Think about it: women are half of the population, and half of our brain power, right? So, as a society, we will do far better if we get the best brains to work on the most pressing problems.”

Combating baked-in biases

The disparities in the field are more than an optics problem; women are missing out on a fast-growing field of intellectually challenging, flexible, and high-paying jobs. And as modern society relies more and more upon technology to make decisions, it becomes increasingly important that the people making the technology reflect the US population in order to avoid replicating existing biases.

Take, for example, a computer program built by Amazon to filter through job applicants. The company’s experimental hiring tool used artificial intelligence to give job candidates scores ranging from one to five stars, and it was trained on prior hiring data. But because a disproportionate number of the people in Amazon’s technical computing jobs were men, the algorithm learned to prioritize the résumés from men over those from women, Reuters reported in 2018. 

Amazon isn’t the only one—a 2024 UNESCO study showed that the large language models underpinning generative AI platforms such as ChatGPT and OpenAI produced “alarming evidence of regressive gender stereotypes.”

The study found that “women were described as working in domestic roles far more often than men—four times as often by one model—and were frequently associated with words like ‘home,’ ‘family,’ and ‘children,’ while male names were linked to ‘business,’ ‘executive,’ ‘salary,’ and ‘career.’

As more people use these generative AI programs, the risk for reinforcing such gender biases (as well as racial and sexual-identity biases) only increases.

“There has recently been a big realization in technology: the importance of designing the solution around the people who are going to use it,” says Bestavros, who is also a William Fairfield Warren Distinguished Professor.

“So, in other words, you don’t sit in your nice offices in Silicon Valley and imagine the target audience for your application,” he says. “If you have not lived it, there is no understanding of how to design the right solution.”

It’s not enough anymore to make something just because you can, Bestavros says. The industry, as a whole, he says, is beginning to focus more on the impact of data and computer science. And so is BU. 

A focus on impact

Bestavros points out that the Faculty of Computing & Data Sciences was built from scratch, not retrofitted from an existing computer science or engineering program. This gave its designers the flexibility to reimagine CDS—what does a data scientist need to learn? What are industry partners seeking from data science graduates? 

“What became very clear is that it is not about the technology,” Bestavros says. “It’s really about the impact of using this technology on all the things that we care about. So, from day one, the message for this program is about impact.”

Kira Goldner, a CDS assistant professor, echoes this ethos. 

“I think we’re very driven by application [of data science], in general, in the faculty research, in the education that we teach to undergraduates and graduates,” she says. “I think that inherently makes it interesting to people, because there are so many different applications, and so you can find yourself in any of them.”

That diversity of interest extends to the faculty, as well, Goldner says. “It’s nice. As somebody who has been in academia for so long, you’re often surrounded by people who do things very similarly to you. Here, we all do data science, but the person next to me is studying the security of various tracking systems and how they affect their users, and across the hall they’re looking into biology applications. It’s just so different and interesting.”

The motivating theory behind CDS is that a good data science education can function much like a good liberal arts education: both can serve as foundations to make any other pursuits more successful. 

“It’s what you need to make an argument. It’s what you need to do critical thinking. It’s what you need to communicate, because today we argue, think, and communicate with data,” Bestavros says. 

By presenting data science as a tool that can be used to work in sports, business, politics, and beyond, CDS appeals to people who may have historically been shut out of or turned off by traditional STEM programs, Bestavros says. And it expands the applicant pool. 

There’s evidence for this in the current CDS cohorts. Amanda Atlas (CDS’27), for example, hopes to join Formula One’s data team once she graduates. Steel plans to join the investment firm BlackRock as a data scientist after she graduates. 

And while their paths may diverge after they leave BU, both women share similar origin stories: they recall being interested in math and computer science from a young age, yet they were often the only women in their high school computer and data science classes. 

“In my Java classes, I distinctly remember being the only girl. I took Java 1 and 2 in high school, and I was the only girl both times,” Atlas says. “I just felt like I had to work 10 times harder to prove that I belonged there.” 

Steel, who came to BU as a math major and changed tack after trying out a CDS introductory course, recalls a similar feeling. 

“In high school when all my teachers for my programming classes were men, I didn’t feel comfortable asking all these questions I had—which were normal questions,” she says. “But I didn’t want them to look down on me, maybe, because I was the only woman in my classes.” 

A “spiral approach” to learning

Steel’s and Atlas’ experiences aren’t unique: girls are still less likely than boys to enroll in foundational computer science courses in high school and earlier. 

This lack of foundational knowledge can end up exacerbating the gender disparity later on. If fewer women are exposed to computer and data science from a young age, they can wind up left out of college programs that demand a certain level of understanding even to begin. 

At BU, CDS solved this problem by making introductory courses that are designed to meet students where they are—prior knowledge or not. Bestavros says the foundational courses use a “spiral approach,” wherein students are given real datasets and then challenged to explore them. Through this exploration, they’ll necessarily need to learn certain math and programming tools—but that learning is in service to a broader, more interesting question. 

Let’s say students get data about the smoking habits of Americans and the lung cancer rates in the country. The professor will then ask whether there’s a correlation between the two sets of data, and explain how to find it. The simplest way is with a math equation, Bestavros says. But doing that equation by hand for datasets that contain millions of data points is time-consuming and arduous. So then, the students will learn how to program a computer to do those equations. This way, they’ll have learned something about statistical correlation and basic computer programming in one fell swoop, all while being guided by their curiosity about what the data contain. 

“We are going to teach math, we’re going to teach coding and programming and all the things that you need to learn in a data science program,” Bestavros says. “But we teach it when you need it, and in order to emphasize how it will help you answer other questions.”

This approach flips on its head the traditional “weed-out” curriculum beloved by many STEM programs, where students encounter some of the most difficult, theoretical courses right off the bat. The ones who can’t hack it, the thinking goes, get weeded out of the program early. 

Within CDS, Steel says, “the intro classes act as a way for anybody—regardless of experience or whether they have seen themselves represented in the field—to learn about data science and see if it is something you would like.”

Steel, who works as a teaching assistant and tutor, says she encounters students from across the University who are taking these foundational courses just to try it out. 

“I hear a lot of people say, ‘I was curious about data science, but I didn’t know if it was right for me’ or ‘I didn’t really know what it was.’ But sitting there and tutoring those students, and seeing the diverse student population that we have coming into these classes, is really unique,” she says. “I don’t really see that elsewhere, and I don’t see it as being as approachable in other colleges.”

This diversity—along gender, race, academic interests—makes for richer classroom discussions, both Steel and Atlas say. 

For Bestavros, CDS’ gender parity is a high point in a long career in computing. He acknowledges that there’s still room for broadening participation, particularly among first-generation college students. But CDS’ success among women is evidence of a long-held theory: we can, and should, do better. 

“To me, I see a problem, but the problem is so hard to fix by a single institution, unless we start going to the root cause,” Bestavros says. “Our program is the first step, and I hope this model becomes a norm. We thought this would work, and now we know it works.”