Human Cognition Colloquium

Human beings can transmit vast knowledge by leveraging a rather unique mechanism of information transfer: language. Human language allows us to communicate which plants are poisonous and that wearing a mask protects ourselves and others, without ever tasting any plants or personally contracting coronavirus. Language is a window into mental representations, but it is a foggy window. Language is vague, the meanings of words depend on context, and speakers are not required to transmit accurate or informative information: We exaggerate, prevaricate, and outright lie in order to accomplish non-informational, socially-motivated goals. How do we see through the foggy window of language to access the abstract knowledge latent in the minds of others?

In this talk, I’ll walk you through work I’ve done over the last several years investigating how we learn from language, how we reason with language, and how we influence other people using language. I gain traction on these complex phenomena through a particular methodology: computational models that formalize precise hypotheses about psychological and linguistic representations, large-scale behavioral experiments to test the predictions of these models, and integrative data analytic techniques that use data to inform psychological theory in a statistically coherent way. I’ll showcase the generativity of the approach through two major case studies: language about abstract category knowledge (so-called generic language) and politeness in language use. In these and other studies, the model-based approach provides insight into the representations that underlie human language understanding, while explaining the variability in human judgments that has perplexed previous attempts to formalize these diverse swaths of human behavior. This work opens the door to understanding precisely how abstract knowledge is transmitted through language, and I’ll close by sharing some ideas about how knowledge transmission through language enables cumulative cultural learning across generations.

Zoom link: https://berkeley.zoom.us/j/92628652633