Yang Tan Collective Alumni Series: Noga Zaslavsky, New York University

Date: Monday, September 28, 2026 Time: 3:00 – 4:00 pm Location: Building 46, Seminar Room 3189 Talk Title: Meaning in mind: Lossy compression in children, bilinguals, and LLMs Abstract: How do diverse minds organize meaning in language? One theory proposes that lossy data compression characterizes word meanings across languages. While this theory has gained substantial empirical support, the evidence comes primarily from monolingual adults, leaving open the question of whether the same principle also guides other populations operating under different learning and social constraints. In this talk, I focus on three such groups: children, bilingual adults, and our recent artificial interlocutors, LLMs. Specifically, I show that: (1) children are not only language learners, but also efficient communicators, achieving near-optimal compression tradeoffs within their developing cognitive capacity; (2) bilinguals, who navigate non-aligned category systems from two different languages, do so while maintaining converged systems that are as efficiently compressed as those of monolinguals; and (3) LLMs, which are not trained for optimizing lossy compression, nonetheless exhibit an emergent inductive bias toward efficiently compressed semantic systems. Taken together, these results suggest that lossy compression underlies meaning in humans and AI, and, more broadly, may be a fundamental ingredient of intelligence. Bio: Noga Zaslavsky is an Assistant Professor of Psychology at New York University and an affiliated faculty member at NYU’s Centers for Data Science and Neural Science. Her research integrates information theory, machine learning, and cognitive science to study the computational principles underlying language, learning, and reasoning in humans and AI. She holds a Ph.D. in Computational Neuroscience from the Hebrew University, and was a visiting graduate student at UC Berkeley, a Yang Tan Collective ICoN Postdoctoral Fellow in the Levy Lab, Fedorenko Lab and Yang Labs at MIT from 2022 through 2023, and served on the faculty of UC Irvine before joining NYU. Zaslavsky was named a 2026 Rising Star by the Association for Psychological Science and her work has been recognized with several awards, including a James S. McDonnell Foundation Understanding Human Cognition Award, two computational modeling awards from the Cognitive Science Society (in 2018 and 2026), and the NeurIPS 2025 Cognitive Interpretability Workshop Best Paper Award.