By Lily Fesler
In June, I attended the second iteration of the Summer Institute for Computational Social Science (SICSS), an intensive two-week program held at Duke that was intended to bring together researchers from across the social science and data science disciplines to learn and discuss topics in computational social science (CSS). Each day, the organizers Chris Bail and Matt Salganik taught mini-lectures on different CSS topics, we split into groups to work on activities together, and a speaker came in to present their research.
Several thoughts have stuck with me since SICSS.
First, we need to continuously evaluate how to conduct ethical research when people are involved. Conducting research in the digital age has opened a lot of possibilities for new exciting research, including using mobile phone data to predict the wealth of households in countries where censuses are cost-prohibitive, using machine learning to reduce jail populations without increasing crime, and using Facebook data to understand how advocacy organizations spread their message. At the same time, it is becoming increasingly important for researchers to be principled in how they conduct experiments and analyze data—even if the companies they work with are not following those same principles. It may have never occurred to Kramer, Guillory, and Hancock that studying emotional contagion by manipulating Facebook users’ news feeds would be considered unethical by many; after all, Facebook conducted experiments like that all the time. However, a company engaging in a practice does not make it ethical, and researchers can and should be held to a higher standard. This does not mean that researchers should never run experiments without explicit consent from participants; without this method, we would have never learned about the extent of employer discrimination against people of color. But, the risks need to be balanced with the benefits.
Matt made the argument that researchers should be following principles based on respecting people, minimizing risks and maximizing benefits, distributing the burdens and benefits of research, and respect for the law and public interest, principles that come from the Belmont Report and the Menlo Report. Following principles is more appropriate than only following the rules in a time when those rules are often outdated, and no longer apply to how researchers are currently collecting and analyzing data.
Second, we should push for more of our resources to be open source and available to more people. SICSS included a relatively small number of participants (30), but all of the lectures and speakers were live streamed to seven other sites (including Helsinki and Cape Town) and almost all of the videos, materials, and talks are still available online for anyone looking for introductions to CSS (which I keep sharing with colleagues who ask me about where to get started in this field). Within my school of education, there is a lot of interest from people with non-computational backgrounds in learning more about CSS, and it is a major step in the right direction to have so many accessible materials that explain the intuition (and often provide some starter code) for new methods that could help us to answer our most pressing research questions.
And third, there’s a lot of power in bringing a group of people together. We had an incredible group of early scholars from across many parts of the globe, studying fields from sociology and political science to network science and biology. It is not always easy during the academic year to connect across disciplinary lines and share research ideas, potential data sources, methods, and tools, but more sharing and feedback makes all of our research better. I was impressed by how much care and planning went into the organization of the institute, which made this experience possible. After the intense academic days filled with lectures and talks, we almost always went out to dinner and drinks together to continue talking about our research and get to know each other better. All I can say is after two short weeks, we all showed up to our final presentations sporting SICSS tattoos (that unfortunately neither Chris nor Matt thought were real, despite our best pranking intentions).
Overall, SICSS was a great experience that asked a lot of hard questions, provided great speakers and resources, and brought together an amazing group of people. I strongly encourage all early scholars who are interested in computational social science to consider attending the institute at Princeton next year, or at one of the many partner sites. And in the meantime, check out all of the materials from this past year!
Lily Fesler is a fourth year PhD student in economics of education at the Stanford Graduate School of Education and an Institute of Education Sciences fellow. Her research interests center around issues of equity in higher education, including barriers to college access as well as classroom bias. She is particularly interested in using computational text analysis to better understand these issues.