“So, what do you study?” Uber drivers, doctors, airplane seat-mates—they all want to know. I often envy my husband, a PhD student in Course X, who can cleanly respond with “chemical engineering.” The interrogator for him often nods politely, acknowledging that these words make sense together.
Me—well, I’m not so easy to explain. As a first-year graduate student at the MIT Media Lab, I live in a constant state of identity crisis. I study computer science but I also study people. The stack of papers on my desk cover topics from game theory to morality to film theory. Like my desk, my studies are messy–they don’t fall into tidy, disciplinarian boxes. And I’m learning to feel comfortable with this.
My undergraduate degrees were in mathematics and computer science from the University of South Carolina. I remember explaining to people that I was learning to write instructions for computers or reassuring them that higher mathematics is more interesting and less traumatizing than their middle school algebra experience. I always expected graduate school to be more (well, a lot more) of the same.
Like many here at MIT, I study artificial intelligence, or “AI.” In fact, MIT recently launched a new initiative making technological advances in the field of artificial intelligence a top priority. But artificial intelligence isn’t the entire story of my work. There’s a missing piece, a piece four years of computer science training did not prepare me for.
I remember thinking to myself, this is so weird, halfway through last semester. I was sitting in a yellow-lit classroom with about ten other students, learning about the difficulties spillover effects pose in experimental design—this was definitely not a course on artificial intelligence.
I was enrolled in MAS.S61: Quantitative Social Science Research Methods, taught by Iyad Rahwan, the AT&T Career Development Professor at the MIT Media Lab, and Nick Obradovich, a research scientist with the Scalable Cooperation research group. I remember thinking, “This is so weird. I’m taking a social science course in grad school. I wouldn’t be learning this if I were studying anywhere else.”
So, how did I, a computer scientist, end up in class with economists, psychologists, and political scientists? The answer dates back to 2015, when Google launched the Google Photos app. Instead of identifying the faces of its black users as it did with lighter-skin users, the app misclassified black users as gorillas. It is one of the earliest and most cited examples of algorithmic bias.
Since then, a growing number of AI systems have exemplified biases such as racism or sexism—phenomena social scientists are uniquely trained to study and an issue I want to fix. This semester, I am taking a course entitled, The Ethics and Governance of AI, taught by Joi Ito, director of the Media Lab, and Jonathan Zittrain, a professor at Harvard Law School. The course is the first of its kind. It is filled with philosophers, lawyers, and computer scientists, each trying to understand artificial intelligence as an actor in our social world.
I’m now learning to ask more questions than “how is this system built?” or “what can we prove about this algorithm?” I’m learning to ask questions like, “what biases are embedded in our systems? How do they behave in the real world? More importantly, “how can we measure them? What tools do we need?”
And sometimes all these new questions lead to me asking—what am I? A computer scientist? A social scientist? I’m learning to be both, and am becoming comfortable with not knowing where one discipline starts and the other ends.
When I first visited the Media Lab as a prospective student, a brilliant PhD student in our group, Morgan, gave me this piece of advice: think about the science you want to do and the science that needs to be done. Well, Morgan, I’m trying. Maybe the title will come later.
Grad Life blog posts offer insights from current MIT graduate students on Slice of MIT.