Coordinating on meaning in communication

Placeholder Show Content


How do we manage to understand each other? Human languages are a powerful solution to this challenging coordination problem. They provide stable, shared expectations about how the words we say correspond to the beliefs and intentions in our heads. However, to handle an ever-changing environment where we constantly face new things to talk about and new partners to talk with, linguistic knowledge must be flexible: we give old words new meaning on the fly. My dissertation investigates the cognitive mechanisms that support this balance between stability and flexibility. Chapter 1 introduces the overarching theoretical framework of communication as a meta-learning problem. Computational models of semantic meaning must explain both the speaker's initial conventional expectations about how words will be understood by novel partners and the dynamics of how these expectations may shift over the course of a particular conversation. Chapter 2 proposes a computational model that formalizes the problem of coordinating on meaning as hierarchical probabilistic inference, which I argue satisfies both of these conditions. Community-level expectations provide a stable prior, and dynamics within an interaction are driven by partner-specific learning. Chapter 3 exploits recent connections between this hierarchical Bayesian framework and continual learning in deep neural networks to propose and evaluate a computationally efficient algorithm implementing this same model at scale in an adaptive neural image-captioning agent. In Chapter 4, I provide an empirical basis for further model development by quantitatively characterizing convention formation behavior in a new corpus of natural-language communication in the classic Tangrams task. By using techniques from natural language processing to examine the (syntactic) structure and (semantic) content of referring expressions, we find that pairs coordinate on equally efficient but increasingly idiosyncratic solutions to the problem of reference. Chapter 5 uses an artificial-language reference game paradigm to test the hypothesis that communicative context systematically shapes which conventions form. Finally, Chapter 6 investigates the generality of the proposed computational mechanisms by examining convention formation in a graphical communication task. Taken together, this line of work builds a computational foundation for a dynamic view of meaning in communication.


Type of resource text
Form electronic resource; remote; computer; online resource
Extent 1 online resource.
Place California
Place [Stanford, California]
Publisher [Stanford University]
Copyright date 2019; ©2019
Publication date 2019; 2019
Issuance monographic
Language English


Author Hawkins, Robert Douberly
Degree supervisor Goodman, Noah
Thesis advisor Goodman, Noah
Thesis advisor Frank, Michael C
Thesis advisor Gweon, Hyowon
Degree committee member Frank, Michael C
Degree committee member Gweon, Hyowon
Associated with Stanford University, Department of Psychology.


Genre Theses
Genre Text

Bibliographic information

Statement of responsibility Robert D. Hawkins.
Note Submitted to the Department of Psychology.
Thesis Thesis Ph.D. Stanford University 2019.
Location electronic resource

Access conditions

© 2019 by Robert Douberly Hawkins
This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).

Also listed in

Loading usage metrics...