New Faculty: Kristin Stephens-Martinez Takes a “Meaning-full” Approach to Data Science

March 09, 2018

Data scientist Kristin Stephens-Martinez's well-ordered mind seeks data of all kinds. She has studied over 4,000 students' wrong answers when they predicted the output of code, collected five-years of her own grad school time management data, tracked her caffeine intake since 2013, and now uses an app to track her son's food, diapers, and naps.

Process optimization comes as naturally to her as breathing. An idea she would like to work on in the future is utilizing the log data from an office hours app at the University of California-Berkeley, where she completed her doctorate in computer science. “Handling student load in office hours and teaching assistant (TA) queues is a control-flow problem,” she said. “They have historical data on what happens and the course context of deadlines. So I want to use that to find a better way to utilize TAs and to decrease student wait times.”

Stephens-Martinez, who is half-Japanese and whose married name is actually Italian, joined Duke’s Computer Science faculty in January as an assistant professor of the practice. She brings a range of experience in software engineering, computer networking, human computer interaction, data science, education, and information visualization.

“I am very interested in making computer science accessible to all students,” she said. This makes her well aligned with the computer science department’s plans here at Duke.

"We are delighted to have Kristin as a colleague,” said Department Chair Pankaj Agarwal. “She will play a critical role in revamping our introductory courses as we broaden their scope and make them appeal to and better serve Trinity students with all majors and interests."

“My research is so intertwined with education itself that it made sense for me to pursue an academic career as a professor of the practice. As a POP, I will have the opportunity to apply the insights of my research and make real changes that improve our courses as a result,” she explains. “That is very appealing to me.“

For her doctoral research, Stephens-Martinez analyzed student educational data from Berkeley’s computer science classes. In particular, she labeled how students got answers wrong when they predicted the output of code and then used the labels to deliver hints to students.

“The sheer breadth of wrong answers was fascinating,” she said, “but there were also a huge number of students who all gave the same wrong answer. I think this points to where students might need hints or that we need to change something in the class itself.”

"My approach to data science is not solely focused on machine learning and algorithms. I believe it's important to understand first the numbers both going into and coming out of the analysis," she explains. "For example, many of the numbers that go into an analysis are some kind of counting, but does the number from the counting mean what you think it means? If I counted how many times a student goes to office hours, what does the number for each student actually mean? Presence in the office hours room, talking to a TA, or having a TA resolve their problem? Once I have an understanding of what is in the data, my goal is to take advantage of those insights by changing or building something in my class."

Stephens-Martinez is co-teaching CompSci 101, Introduction to Computer Science, this semester with Owen Astrachan.

Additionally, she is passionate about undergraduate research and diversity in computing science—two themes that are integral to the department’s vision. "I've led the computer science women groups both in undergrad and grad, and mentored many women to pursue both research and industry paths. I hope I can serve as another data point showing other women that we can pursue computer science and succeed."

“Mentoring is very important to me even now as a professor. One of the reasons I came to Duke Computer Science is the deep bench of high profile mentors at the professor of the practice level. Owen Astrachan, Susan Rodger and Jeff Forbes have a lot of experience as teachers and grant-funded researchers that I can draw on. I didn’t see that at other institutions,” she said.

Stephens-Martinez not only cares about being mentored, she believes in giving back and mentoring others. "Whenever I see someone who looks like they need a chat, I try to ask if they'd like to go for a coffee and a walk with me," she said. Besides these casual mentoring sessions, she mentored 10 graduate and 13 undergraduate students at Berkeley during her graduate student career.

“One of my favorite experiences with mentoring was with a student who was not doing well in a section I TA’d. I reached out to him when I saw he was struggling and told him to calculate what he needed to do to get the grade he wanted. Then we made a strategy to help him reach his goal,” she explains.  “Getting the grade was the end goal, but I had him think of reaching that goal as a marathon. We worked out a strategy for him to ‘train’ by practicing problems while timing himself to create stress that mimicked taking a real exam. This conditioning helped him perform well when it came time for his exam. In the end he mastered the material and got the grade he wanted on the final.”

Kristin Stephens-Martinez