Data is the bread and butter of academic research. But when it comes to the major digital platforms of the day, where many people are spending the bulk of their time, researchers do not have access to robust, independent tools that help them study or understand the conversations taking place.
Enter The Accelerator, an initiative at the Princeton School of Public and International Affairs to build public-interest infrastructure that allows researchers to better understand the “information environment” – or the places and platforms where information is exchanged.
When that environment was largely analog – with ideas exchanged in newspapers or on television – it was easier to catalog and understand. Today, the conversations happening on TikTok or YouTube are much harder for researchers to access, particularly because many of the tools that track them are designed for brands and marketers. The companies themselves are also not a reliable source of complete data, which is critical for holding them accountable.
“Right now, researchers are trying to understand a fast-moving, high-stakes domain using tools that mostly weren’t designed for scientific work,” said Jacob Shapiro, the John Foster Dulles Professor in International Affairs and vice dean for strategic initiatives at Princeton SPIA, who leads The Accelerator. “The result is that policy debates often rely on anecdotes, leaks, or correlational studies that can’t answer causal questions.”
The consequences of that go beyond ivory towers, as researchers look for ways to study whether social media harms teenagers, for example, or how false information spreads during crises.
The Accelerator addresses that challenge by developing secure analytic tools and data collections that meet academic review standards. While it is nearly impossible to analyze all of a platform’s content, the team is collecting samples of activity across Telegram and YouTube, utilizing novel methodologies meant to be representative of the majority of each platform’s activity. Instead of wrestling with restrictive application programming interfaces (APIs), building time-consuming scrapers, or manipulating tools built for advertisers, the goal is for researchers to quickly access data and surface insights relevant to their research areas.
For example, in the leadup to the Iran war, The Accelerator captured a high volume of Farsi messages on Telegram – a platform banned in Iran but nonetheless used there – that could help researchers understand how people on the ground were discussing the threat of war.
“Our goal is to shorten the distance between a research question and a meaningful answer. When researchers aren’t spending weeks or months just getting data into usable shape, they can take on bigger, more ambitious questions," Shapiro said.
The Accelerator was built by a multidisciplinary team of engineers, data scientists, social scientists, and policy experts with the help of a global consortium of partner organizations. Funders include Princeton SPIA, private sector companies, major foundations, and private donors. Nearly 40 users beta-tested its tools last year, and the first research cohort of more than a dozen research teams is now diving in.
The use cases are as varied as academic research itself. Faculty at the Paris School of Economics are working with Rutgers School of Public Health and Princeton to generate causal evidence on social media’s mental health impacts. A Princeton research scientist is collaborating with the Rutgers Center for Gambling Studies to analyze the rise of unregulated online gambling.
At Princeton SPIA, the Empirical Studies of Conflict Project, which Shapiro co-founded, is using Accelerator resources to measure engagement with foreign-influence campaigns, assess threats posed by generative artificial intelligence, and evaluate how information spreads across platforms. The Center for Information Technology Policy is examining how short-form video content influences overall cell-phone use among young adults.
Building these tools is not easy. Developers couldn’t simply draw random samples of digital platforms using existing tools. Because a tiny fraction of content often drives most of the engagement they had to build structured, stratified approaches that reflect how attention actually works online.
The team was also committed to building trustworthy tools, which means adhering to ethical standards such as privacy-preserving storage, role-based access, and audit trails, as well as adhering to platform terms-of-service. That’s why they were able to collate YouTube descriptions and comments, but not the videos themselves, which are protected by copyright. While that is limiting, it also ensures that research conducted through the Accelerator will meet the standards of university institutional review boards.
“Most commercially available tools are not meant for academic research,” said Kristen DeCaires, director of operations and research development at The Accelerator. “They are often designed with another market in mind; they use data collected in a manner that would never pass through an IRB-review board or hold up to common publication standards - they’re simply not tailored for these purposes.”
The team continues to build out the tools and plans to add additional platforms. It's also working to secure additional resources to meet the high levels of cloud usage and computing power needed to house such large datasets and make it easy for scholars to work with them.
“The combination of a strong team, a committed consortium, and broad-based support is what’s made it possible to build something this ambitious and keep it grounded in the public good,” Shapiro said.
Image Credit: Shutterstock