ML-NYC Speaker Series and Happy Hour: Matt Salganik

The ML in NYC Speaker Series + Happy Hour is excited to host Matt Salganik from Princeton University as our April speaker! His talk will take place Monday, April 20th at 4pm at the Flatiron Institute. As always, there will be a reception afterward for all attendees.

NOTE: Registration will close on April 18th. Please be sure to register beforehand! Registration is required for entrance.

Title: Life Trajectories and Life Chances: New Approaches from Population Registries and AI

Abstract: How predictable is a human life? This talk presents two projects that approach this question from opposite directions using population-scale registry data from the Netherlands. In the first project, we try to push prediction accuracy forward. We develop the "book of life" approach, which represents life course data as textual narrative rather than tabular features, and fine-tune an LLM on millions of these narratives. We then run a head-to-head comparison against standard tabular methods for predicting life outcomes such as having a child. In the second project, we try to estimate a bound on predictive accuracy, even given infinite data and the best possible model. Using 200,000 twin pairs — who share everything about their birth circumstances, measured and unmeasured — we characterize the best possible prediction function from birth circumstances to life outcomes, without constructing it. This approach reveals how much of life is determined at birth versus shaped by what comes after, and establishes bounds that no model, however sophisticated, can exceed. I see these projects as examples of a growing intersection between social science and machine learning — where old questions and new data and methods speak to both communities.

Bio: Matthew Salganik is the Alexander Stewart 1886 Professor of Sociology at Princeton University and former director of Princeton’s Center for Information Technology Policy. He is also affiliated with several interdisciplinary initiatives at Princeton, including the Princeton AI Lab and Princeton Precision Health. His research spans computational social science, social networks, and the use of digital data and technologies in social research. He is the author of the award-winning book Bit by Bit: Social Research in the Digital Age. His work has received honors from the American Sociological Association, the American Statistical Association, and the American Association for Public Opinion Research, and he has also received Princeton’s Commendation for Outstanding Teaching. A former Fulbright fellow, he earned his Ph.D. in sociology from Columbia University and his M.A. in sociology from Cornell University.

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher ML-NYC Speaker Series and Happy Hour