Kathryn Bjorklund Attends Computational Social Science Program - Barcelona
In October, PhD student Kathryn Bjorklund attended the Summer Institute for Computational Social Science program hosted by the University of Barcelona. The program, which brings together graduate students, postdoctoral researchers, and beginning faculty interested in computational social science, involves lectures, problem sets, and participant-led research projects.
The lectures covered topics like text as data, images as data, APIs, web scraping, digital field experiments, network analysis, machine learning, and ethics. Researchers Andreu Casas and Chico Camargo spoke about their work on different studies, both of which involved YouTube—one on the moderation of political content and another on the platform’s recommendation algorithm.
The program brought together about 20 participants working in different parts of the European continent. The topics people were studying were equally diverse: human interactions with climate services, patterns of human mobility and mosquito transport, social movements, democratic backsliding, and more. Toward the end of the program, participants split into groups to create pilot studies. Kathryn and her group used the Global Database of Events, Language, and Tone (GDELT) database and survey data to measure the sentiment of news coverage about events occurring at different distances from the media outlet's home base.
Following the program, Kathryn attended a Computational Social Science Conference organized by the Barcelona Supercomputing Center, focusing on innovative methods, research workflows, and data stewardship. David Lazer delivered the keynote speech and emphasized the need for developing a science of human behavior on the internet, given that an increasing fraction of life takes place online. He highlighted the challenges posed by the shutdown of APIs from platforms like Twitter, likening it to losing the single lamppost in an otherwise dark and expansive parking lot. He also outlined four alternative models for data collection but ultimately only vouched for one: volunteer-sourced data with shared infrastructure.
Other speakers included the likes of David Garcia, who explored in his work how opinions form, spread, and change within a simulated society of AI agents, and Hannes Mueller, who spoke about his tool ConflictForecast, which ingests news media and employs topic modeling to forecast crises and conflicts.
Kathryn also visited Barcelona's supercomputer MareNostrum 5, which boasts the computing power of about 380,000 laptops and has around 400 petabytes of storage. She saw the water systems used to cool down the machinery, a sample accelerated partition node, and the old cathedral where it used to be housed. The tour concluded with a stroll down a hall of supercomputer models past, allowing for reflection on how we use these technologies to tackle global challenges and how, in turn, these technologies influence us.