Online-Workshop on Rational Speech Act Theory and its Applications

Abstract

The Rational Speech Acts (RSA) framework models speakers’ choices of linguistic expressions and addressees’ interpretations as the outcome of recursive, utility-based reasoning in communicative interaction. Speakers balance expected communicative success against production costs, while listeners reason about speakers’ goals and beliefs. A central assumption of RSA is that interlocutors are Bayesian reasoners with bounded rationality, whose inferences are shaped by probabilistic beliefs, contextual uncertainty, and cognitive constraints.

Over the past decade, RSA has become one of the most influential and actively developed approaches at the interface of formal semantics, pragmatics, and cognitive modeling. It has provided a unified perspective on a wide range of phenomena, including reference, scalar and gradable expressions, presupposition, question–answer relevance, over- and under-informativity, and communicative adaptation.

This online workshop brings together established researchers and early-career scholars working on RSA and related Bayesian models of meaning and communication. The talks span experimental, formal, and computational perspectives, including probabilistic models of semantic interpretation, pragmatic learning and adaptation, multi-agent and reinforcement-learning approaches to communication, and the role of utility, relevance, and context in shaping linguistic behavior. By highlighting both empirical results and theoretical developments, the workshop aims to promote dialogue across linguistics, philosophy of language, and cognitive science.

Invited speakers

• Benjamin Spector: https://sites.google.com/site/bspectorpage/
• Kristina Kobrock: https://kristinakobrock.github.io/
• Alexandra Mayn: https://www.uni-saarland.de/lehrstuhl/demberg/members/alexandra.html
• Judith Tonhauser: https://judith-tonhauser.github.io/
• Polina Tsvilodub: https://polina-tsvilodub.github.io/home/

Local Organisation

• Anton Benz (ZAS Berlin)
• Hening Wang (U Tübingen)

Scientific advisors

(the members of the LMBayes project)

• Anton Benz
• Michael Franke
• Davit Gogolashvili
• Ari Joshi
• Uli Sauerland
• Vladimir Spokoiny
• Karsten Tabelow
• Hening Wang

The workshop topic is a key focus of the new project “LMBayes” (Linguistic Meaning and Bayesian Modelling), which is based at ZAS Berlin in collaboration with WIAS (Weierstrass Institute for Applied Analysis and Stochastics) Berlin and the University of Tübingen (https://www.leibniz-zas.de/de/forschung/forschungsbereiche/semantik-pragmatik/lmbayes). 

Programme 

Date: Friday, 09.01.2026

10:00 – 10:10 Introduction
10:10 – 11:00 Benjamin Spector (Paris): New evidence for a Bayesian approach to gradable expressions
11:10 – 12:00 Kristina Kobrock (U Osnabrück): The role of pragmatic reasoning in the interaction‑based emergence of referential communication: Integrating the Rational Speech Acts framework into a multi‑agent reinforcement learning model
12:10 – 13:00 Alexandra Mayn (U Saarland): Modelling Listeners' Adaptation to Speakers' Pragmatic Competence

Break

14:00 – 14:50 Judith Tonhauser (U Stuttgart): Presuppositions as inferences about private speaker assumptions
15:00 – 15:50 Polina Tsvilodub (U Tübingen): Modelling relevant answer to polar questions
 

Please contact benz@leibniz-zas.de for the ZOOM link.