About

Charles Yang is a professor at the University of Pennsylvania, where he specializes in linguistics, computer science, and psychology, with a focus on learning by humans and machines. He studied computer science at the MIT AI Lab and has held faculty positions at Yale before moving to Penn. At Penn, he teaches and conducts research in his interdisciplinary fields, serving as the Chair of the linguistics department and having been the Director of the Program in Cognitive Science since 2011. He co-directs the Integrated Language Science and Technology group with John Trueswell. Yang has authored several books, including 'The Price of Linguistic Productivity,' which was awarded the Leonard Bloomfield Award from the Linguistic Society of America. His work has been supported by fellowships from the National Science Foundation and the John Simon Guggenheim Foundation.

Research topics

• Artificial Intelligence
• Computer Science
• Natural Language Processing
• Linguistics
• Psychology
• Economics
• History
• Macroeconomics
• Geography
• Philosophy
• Statistics
• Mathematics
• Cognitive psychology

Selected publications

• A simple threshold captures the social learning of conventions

  Proceedings of the National Academy of Sciences · 2026-04-22

  articleOpen accessSenior authorCorresponding

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the tolerance principle (TP), a parameter-free equation developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer nonlinguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• Close enough isn’t good enough in word learning: successful cross-situational word mappings are semantically independent of previous mappings

  Journal of Memory and Language · 2025-09-08

  articleOpen access
  PublisherOA PDFDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-11-17

  articleOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-11-19

  articleOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-04-03

  preprintOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• Lead the Way: Halide Perovskites as Next-Generation Triplet Sensitizers for Photon Upconversion

  Chemical Reviews · 2025-11-04 · 7 citations

  review

  Photon upconversion, the process of converting low-energy photons to higher energy ones, shows promise for applications in solar energy, photocatalysis, biomedicine, and additive manufacturing. In triplet-triplet annihilation (TTA), incident low-energy photons populate metastable spin-triplet states that annihilate to generate high-energy emissive spin-singlet states. Thus, TTA-based photon upconversion (TTA-UC) can operate efficiently under incoherent and low-intensity excitation, such as sunlight. In this Review, we discuss the recent emergence of halide perovskite-based materials as potent triplet sensitizers for a variety of applications. Due to their strong and tunable absorption and high defect tolerance, perovskite materials ranging from nanocrystalline to bulk semiconductors enable efficient TTA-UC in both solution and solid-state systems. After introducing the TTA-UC process and giving a brief overview of its beginnings, we first consider TTA-UC systems based on perovskite nanocrystals and low-dimensional perovskite-inspired materials and the achievements that have been made in those areas. We then focus on the mechanism of bulk perovskite-sensitized TTA-UC, the impact the underlying structure holds, and review the current challenges in perovskite-sensitized solid-state UC and outline future research directions to unlock the full potential of TTA-UC in practical applications.

  PublisherDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-11-04

  articleOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-11-04

  articleOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherDOI

• A Simple Threshold Captures the Social Learning of Conventions

  2025-04-04

  preprintOpen accessSenior author

  A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction. Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks. Across experiments, participants' choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize. We effectively estimate this threshold using the Tolerance Principle (TP), a parameter-free equation first developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions. The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer non-linguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants. These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.

  PublisherOA PDFDOI

• Phonological Regularity and Breakdown. An Account of Vowel Length Leveling in Middle English

  2024-01-01

  book-chapter1st authorCorresponding
  PublisherDOI

Frequent coauthors

• Julie Anne Legate

  15 shared

• David Pesetsky

  10 shared

• Jordan Kodner

  8 shared

• Sarah R. Payne

  7 shared

• Erwin Chan

  7 shared

• Deniz Beser

  6 shared

• Mitchell P. Marcus

  Philadelphia University

  5 shared

• Jacob A Lichtefeld

  4 shared

Labs

• Department of Linguistics, University of PennsylvaniaPI

Education

• Ph.D., Language acquisition, language change, computational linguistics, morphology, psycholinguistics

  MIT

  2000