Tamar Mitts: New Tools to Understand Old Problems
Tamar Mitts — whose research uses big data, machine learning, and text analysis to study conflict, radicalization and violent extremism — joined SIPA’s faculty as an assistant professor in fall 2018. Mitts holds a PhD in political science from Columbia University and taught previously at the University of Michigan. She is also a faculty member at Columbia’s Data Science Institute. Her current research examines the behavior of Islamic State supporters on social media, drawing on new data of over a million users linked to the extremist group on Twitter.
The interview has been edited for length and clarity.
How did you get interested in studying radicalization and violent extremism?
I grew up in Israel and served in the Israeli Military Intelligence during the Second Intifada. This experience had a profound impact on me and I realized how important it is to understand why young individuals might risk their lives to join groups that perpetrated violence against civilians. What about their lives in a conflict zone attracted them to the extremist world? How might armed groups exploit these horrible situations to attract supporters? I’ve been interested in these questions for years.
Could you describe the interdisciplinary nature of your research?
My current research, focusing on the ways in which groups like the Islamic State mobilize supporters via online platforms, draws on multiple disciplines. First, we need to understand the recruitment strategies of armed groups and the psychological process of radicalization. In addition, since the online world has been playing a major role in this space, there are exponentially more data today that we can use to understand radicalization and recruitment into violent extremist groups. It is not enough only to have qualitative policy expertise on the phenomenon, we also need excellent data analysis skills and the ability to work with big data.
What is it like to conduct research across disciplines here at Columbia University?
One of the most exciting things I’ll be doing this year is to teach a course on data science for public policy. In this class, we will bring students from SIPA and the Data Science Institute to work together on cutting-edge policy problems with large amounts of data. Throughout the semester, students will have the opportunity to analyze real-world data sets on a broad range of policy topics, including, for example, data on Russian trolls disseminating misinformation on social media, data on Islamic State recruitment propaganda on the Internet, and granular information on natural disasters that can facilitate preparedness for future hazards. A lot of exciting things are happening at the intersection of public policy and data science here at Columbia, and I am thrilled to be a part of it.
How do you use online data to study the Islamic State?
Since 2011, the Islamic State has been extensively using different social media platforms to communicate with its recruits, attract supporters, govern different areas, and disseminate propaganda. With all of these activities online, I saw an opportunity to learn about ISIS through the lens of the virtual world.
For example, during the past few years I was able to collect information on Islamic State networks on Twitter, capturing data on the online behavior of over 1.6 million users who followed accounts that disseminated ISIS propaganda on the platform. I was able to observe how these users communicated with each other, when they were exposed to propaganda, and how different events on the ground during the Syrian civil war shaped their views on ISIS. I obtained over 100 million tweets produced by these users over the course of a few years.
In one of my studies, I focus on ISIS sympathizers in Western Europe, trying to understand how day-to-day life experiences in the offline world might shape online behavior in ISIS’s social networks. Using new machine learning tools to find the geographic location of social media users, I found that ISIS supporters located in areas with high levels of hostility towards minority populations were significantly more likely to show signs of online radicalization. The ability to connect the online and offline worlds with vast amounts of data allows us to paint a picture that we would not have been able to observe without drawing on data science methods.
What exactly is machine learning? How do you use this technology in your research?
Put simply, machine learning is a just way to teach a computer to recognize patterns in data and make predictions based on those patterns. Machine learning algorithms are used everywhere today, from detecting faces in photos, to predicting the shortest route to a destination, to automatically translating text.
In another project in collaboration with scholars at the University of California San Diego, we draw on video-as-data object detection and audio-to-text transcription algorithms to identify recruitment messages in thousands of Islamic State propaganda materials that were disseminated online over the past several years. To measure how propaganda messages influence potential supporters, we use my ISIS Twitter database to detect when and where these propaganda were disseminated, and how those who were exposed to propaganda subsequently reacted. These are important national security questions that are hard to answer without data science tools.
How has other research you have conducted seek to address the issue of radicalization?
In another research project, I study what might be done to deradicalize individuals who are on the path towards extremism. One of the challenges with studying countering violent extremism (CVE) is to understand its effectiveness. In order to shed light on this problem in the context of the United States, I draw on over one hundred CVE events organized by the U.S. government and match them data on the online behavior of ISIS sympathizers in these areas. I find that CVE activities led to a significant decrease in online pro-ISIS chatter among ISIS sympathizers, but am still investigating whether this decrease was due to deradicalization or self-censoring as a result of greater awareness of government monitoring.
Looking ahead, what is the next big question in your field of study?
We are all aware that radicalization is happening not only in the Islamist context, but also among far right and neo-Nazi groups. In future work I plan to study radicalization and extremism in these circles as well. In addition, since rapidly-evolving media technologies are shaping various social and political processes around the globe, I plan to study how the online world influences behavior in the offline world, and how democracy and social order can be preserved in this new media environment. We live in an era where change is constant and information abundant. Data science allows us to study old questions with new tools. I think it is an exciting frontier that I hope many more will embrace because I believe this is where the future is.
