Four years ago we did something that hadn’t been done before at the UC Berkeley School of Information. We created a Master of Information and Data Science program designed to attract the best and the brightest, regardless of where they lived. It would be interdisciplinary, incorporating not just math and modeling, but also social science and policy. It would have great professors, small class sizes, and plenty of networking opportunities. But it would be different from other data science programs in one important way: it would be fully online.
By eliminating physical limitations, we aimed to reach more outstanding students in more places—especially those who weren’t in a position to move to Berkeley for two years. We partnered with education technology company, 2U, to build an innovative digital curriculum. And then we put it to the test.
Right away, we saw the power of the digital model. Our first class was a diverse group of students from eight states across the country, with backgrounds ranging from economics and astronomy to neuroscience and business operations. Many of these students already held advanced degrees—MBAs, JDs, and PhDs—and had extensive work experience to boot.
We also noticed a familiar trend: women were underrepresented in these early cohorts. In our first class, they made up only 20 percent of the student body. This kind of gender imbalance has plagued the STEM fields for decades. Fortunately the recruitment and retention data collected by 2U provided a possible answer and path forward. We saw first that we were retaining and graduating women at the same rate as men—so the problem was not a lack of success in the program. We could also see that many women were intrigued by our program, and committed enough to begin their applications. But at some point before clicking “submit” these women were dropping out.
To fix this problem—and, we hoped, help the best candidates make it through our rigorous application—we created a “bridge course.” The goal was to help potential students achieve competence in Python, the premier high-level programming language for data science, to bridge the gap in their knowledge.
We hoped that the bridge would help us widen our pool of candidates. But we were especially delighted to see the bridge course immediately filled with women. And the following year, the percentage of women participating in our data science program shot up by 33 percent.
It’s the kind of insight that’s so simple that it almost seems obvious: by helping applicants fill a gap in their skills, we could welcome capable women who had shied away from completing the application to our program, likely because of their lack of programming background.
Lisa Barcelo, one of the students who enrolled in our bridge course, put it best. “As a woman in the workplace, it’s easy to be deferential when presented with new challenges and to say, ‘I’m not sure or I don’t want to assume that my answer is correct,’” she told me. “This course helped me to not only develop my skills, but also my confidence. I learned that I could find solutions to complex problems both in my academic endeavors and at work. That’s the most powerful thing.”
There’s a lot more work to do, on the part of schools, businesses, and the public sector, to close the gender gap in data science and other STEM fields. But by using technology to broaden our applicant pool, and data analysis to identify and fix breaks in our system, we’ve created a more inclusive program—one where women are represented at a rate more than 50 percent above the national average.
As educators, we have an opportunity, and an obligation, to use new tools to reach new students. In the same way that our students—through trial, error, and insight—learn to master their subjects, we also find new ways to expand the reach of our teaching. When we create more inclusive programming, the whole world is better off for it. Just look at the data.