As artificial intelligence continues to provide unparalleled and unprecedented access to information and analytic tools, Jonathan Mummolo believes the work of Princeton University’s Data-Driven Social Science (DDSS) initiative is more important than ever.
“As with any transformative research technology, AI presents great opportunities, but also great risks,” Mummolo said. “DDSS is focused on providing resources to Princeton researchers to help harness these tools in social science research while maintaining care and rigor.”
Mummolo, an associate professor of politics and public affairs whose research studies bureaucratic politics and political behavior, was named the interim director of DDSS in August. Under his leadership, Mummolo is positioning DDSS and its team of research software engineers to help Princeton’s social science community harness AI to answer important social questions with attention to the unique challenges social science presents, including data privacy, replicability, and computational obstacles associated with massive data sets such as video archives.
Housed under the Princeton Statistics and Data Science pillar of the newly formed academic unit Data and Intelligent Systems (DaIS), DDSS supports social science research through one-on-one consultations, workshops, symposia, research grants, and the development of computational tools that take advantage of Princeton’s existing infrastructure while meeting the rapidly evolving needs of the field.
DDSS has developed a variety of computational tools for Princeton faculty and graduate students, including Blackfish, a user-friendly interface for running powerful open-source AI models at no cost — without the need to be a programmer or send data to an outside company. Mummolo believes this tool will be increasingly in demand for social scientists who may be analyzing sensitive data that cannot be shared with AI companies, or seeking to avoid “black box” large language models in their analysis.
“Off-the-shelf AI tools are often inappropriate for cutting-edge research,” Mummolo said. “With the tools DDSS has developed, social scientists can use powerful and transparent approaches without ever sending their data to a private company.”
Another original tool, TigerFlow, automates large, multi-step data-processing jobs — like converting millions of documents or thousands of hours of video into machine-readable data — so researchers don't have to manage every step or spend months learning to manage a computing cluster.
The initiative’s restricted cloud environment streamlines the process of analyzing sensitive data at scale, with voter and commercial data that are research-ready, saving researchers weeks to months of coding time.
"As AI accelerates the pace of discovery, the deliberate strategy of DDSS to build a strong technical team over time will remain an invaluable source of expertise — enabling the impactful, responsible integration of AI into the social science workflow," said Lori Bougher, DaIS director, research and strategy. “DDSS builds the infrastructure so social science researchers can focus on the questions that matter, working at unprecedented speed and scale.”
DDSS will also host guest speakers from Google DeepMind, Anthropic, and OpenAI throughout the year to demonstrate evolving applications of computational social science.
“This is an incredibly exciting moment to be doing social science,” Mummolo said. “DDSS is here to help Princeton researchers make the most of it.”