Machine Learning for Social Science - Hanna Wallach

by Nihar Patil

For years, technologists have questioned to what extent computers can understand and predict human behavior. Hanna Wallach, a Senior Researcher at Microsoft, is one of these technologists, but she doesn’t stop at questioning - she finds answers. She uses “machine learning methods to uncover insights about the ways in which people interact and work together.” Her area of expertise centers around a seemingly dichotomous subject area – computational social science – but more on that later.

 

Dr. Wallach’s work is, by nature, highly interdisciplinary – but this wasn’t always the case. As a teenager, she held an intense desire to study the social sciences but was advised against this by her school. Dr. Wallach explained,

“In the United Kingdom, people specialize very early. At age 15, I stopped taking classes in all subjects apart from math, physics, and something called 'design technology.' I wasn't able to study the social sciences because I was studying the subjects that would let me pursue a career in engineering.”

It’s important to distinguish that engineering in the United Kingdom, where Dr. Wallach is from, has its own distinct path from computer science. “Engineering” includes mechanical engineering, structural engineering, electrical engineering, and electronic engineering – but not computer science. This begs the question – how did Dr. Wallach make the transition from wanting to study the social sciences to studying engineering to being a computer scientist?

“I went from never having touched a computer to working full time with machine learning. It’s funny – in high school I thought that computers were primarily for typing documents and that they just weren’t interesting. Then, during my 'gap year' between high school and college, someone asked me to design a website. I asked, ‘What’s a website?’ They told me about the Internet and gave me a book on HTML. This changed everything. Almost immediately, I became obsessed with computers.”

Despite this revelation, Dr. Wallach started college as an engineering student. However, at the end of her first year, she switched to computer science. When Dr. Wallach entered the computer science program, she was behind. She had been specializing in engineering, so she did not have the same level of computer science knowledge as the other students. She was undeterred. Dr. Wallach was not afraid to make the switch because she knew computer science was the right path for her. However, she still felt that she'd missed out by not studying the social sciences.

“That same year, I read an article profiling the research of one of my now colleagues. This article showed me that I could use math and computers to study social phenomena. Until that point, I thought that I had to choose between computers and the social sciences. The article made me realize that I could effectively contribute to the the social sciences using the skills I was learning as a computer scientist.”

After her undergraduate degree, Dr. Wallach completed an MSc in cognitive science, before specializing in machine learning for her Ph.D. She now collaborates with political scientists, sociologists, statisticians, and computer scientists to analyze the structure, content, and dynamics of social processes. Social processes are incredibly complex, filled with biases, irrational and rational decision making, emotions, and a wide range of other human factors. Dr. Wallach and other pioneers in computational social science use digitized data and computational methods to study social processes through mathematical and analytical eyes. Dr. Wallach's research focuses on uncovering patterns in observed data, such as news articles, meeting transcripts, public records requests, and emails. “For example, a political scientist might want to use congressional votes and bill text to understand when and why senators' voting patterns deviate from what would be expected from their party affiliations.” From this example alone, it’s easy to see that the applications of machine learning in the social sciences are vast. For our machine learning junkie readers, you can view Dr. Wallach’s work here. Dr. Wallach encourages computer scientists and social scientists to “push the boundaries by working together to tackle fundamental questions. Computational social science won’t advance unless we have strong researchers with complementary backgrounds.”

 

In hearing about Dr. Wallach’s journey and her decision to focus on machine learning for the social sciences, I was curious about any obstacles she had faced.

“Specializing so early [in the UK system] meant that there was no 'general education' component to my college studies, so I've had to teach myself about the social sciences. Additionally, it's not easy to be a woman in computer science. I've had to work hard to be taken seriously as a computer scientist, and I've encountered both implicit and explicit biases.”

Although some students complain about gen ed courses, Dr. Wallach’s perspective sheds light on the unique privilege that general education curricula provide. Regarding her second point, I was taken aback to hear that one of the researchers responsible for kickstarting an entire field of study – who has contributed to open source libraries, was one of Glamour magazine’s “35 Women Under 35 Who Are Changing the Tech Industry,” and has been featured in several Linux magazines – was “not taken seriously” because of her gender. Fortunately, Dr. Wallach plays an active role in promoting and supporting women in computing. She co-founded the WiML (Women in Machine Learning) Workshop, which has grown from around 100 attendees in 2006 to over 600 attendees in 2016. She also co-founded the GNOME Women's Summer Outreach Program and the Debian Women Project. Computer science should be for everyone – it’s heartening to see Dr. Wallach championing women to join and stay in the field. 

“I often see women and minorities drop out of computer science because they don’t believe that they’re good enough. It's easy to suffer from imposter syndrome.”

Throughout the interview, Dr. Wallach mentioned key advice for undergraduates. She recommended that all undergraduates work to develop “grit” – a combination of passion and perseverance – as well as a “growth mindset.”

“Things aren't always easy. But, if you persevere, they will become easier in the long run – if you invest your time and your energy, you will improve. It's also important to accept mistakes and failures as important steps on the path to success.”

Dr. Wallach’s journey as a computer scientist and one of the leading researchers in computational social science is telling of her perseverance, her love for her research, and her vision for using computers to answer fundamental questions about society.


Nihar Patil

Nihar is a sophomore majoring in Networks and Social Systems Engineering, and is the founder of PTR. You’ll either find him dancing with his friends in a giant carrot costume or playing guitar and writing code.