Weaponized Feeds and Text as Data: Notes from IC2S2 2017

We just returned from the 3rd International Conference of Computational Social Science, hosted by GESIS-Leibniz Institute for the Social Sciences in Cologne, Germany. It was an amazing conference -- really well organized by the incredible local team -- with a bunch of really interesting talks across the three days. It was inspiring to meet so many people passionate about the same stuff we are!Some of our highlights were…Ciro Cattuto gave a great keynote on high dimension networks from wearable sensors. The advances in research using sensors over the last few years are exciting, and the applications are wide-ranging, including helping us understand the spread of infection in hospitals and schools. As they get cheaper, become self-contained systems and have a longer battery life, they can be deployed in new research settings, include emergency and disaster settings. Watch it here:https://youtu.be/RjJoxFVnpHwKathleen M Carley gave a slightly terrifying but really interesting keynote on social media influence and bots. Estimates suggest 5 percent of Twitter accounts are bots -- that’s 230 million users! -- and that 24 percent of all tweets are generated by bots. She showed examples of how removing bots from networks dramatically changed the property of networks, which is an important point for network analysts, and also showed examples of how bot networks have changed behavior on Twitter -- from leading people to a charity app which launders money for ISIS, to encouraging young men in Ukraine to form an online community and then increasingly weaponizing their feeds. And, of course, they are sharing all this fake news we’re always hearing about. Anyone who thinks bots are harmless needs to watch this!https://youtu.be/9tWiVvx2qLsJustin Grimmer’s keynote (that's him at top), which was sponsored by SAGE Publishing, talked about the “text as data” revolution taking over the social sciences, and how we might make causal inferences from text. In addition to presenting a framework for causal interventions with text, he makes a strong case for taking a sequential (inductive) approach to social science to build theory with successive experiments, standing in contrast to those calling for pre-analysis plans in political science. Watch it here:https://youtu.be/3vNFT-ck5NEJeff Hancock gave a brilliant keynote on research he’s done following the infamous Facebook news feed manipulation research he was involved in. After analyzing the hate mail (!) he was sent following the publication of that research, he realized that where they’d gone wrong when considering the ethics of that research was that they didn’t understand enough about people’s expectations and understanding of how Facebook’s news feed worked. Using metaphors and a cool new tool I hadn’t heard of called wiki surveys, they uncovered the folk theories users hold to explain how social feeds work on platforms. Knowing what users think about the platforms they are using can help us design future studies… without the hate mail. Watch it here:https://youtu.be/h_n6EZ3VhvsThe parallel sessions were also great, with papers ranging from bots on Wikipedia, to the motivations of crowdworkers, to using Google searches for ultrasounds to predict abortions, and much more.And to top it all off, we did our double act and presented the findings of our white paper and shared some of the things SAGE are doing to support the acceleration of computational social science, including the recent launch of SAGE Campus - online data science courses for social scientists.

All in all, a brilliant couple of days. Already looking forward to next year in Evanston, Illinois!

*** Katie Metzler is head of methods innovation at SAGE Publishing. Ian Mulvany is head of product innovation at SAGE. They tweet @KMetzlerSAGE and @IanMulvany