Description. Game theoretic models have found diverse applications in linguistics. These range from rational (probabilistic) language use and language evolution up to language emergence and the establishment of conventions. The goal of this project is to give an overview of these areas and to provide students with the necessary background to conduct their own inquiries in an area of their interest.
We will look at language from two different, though related, points of view: (i) the negotiation of language meaning in particular discourses using standard game theory, and (ii) the evolution of meaning using evolutionary game theory. As for (i) we will motivate Gricean principles of language use, and for (ii) we will investigate how some particular features of natural languages can evolve within and across populations under consideration of linguistic pressures such as learnability and expressivity.
Organization & schedule The project is organized in two parts. The first part will consist of lectures starting January 16th/17th. The second part will be spent on individual or group (sub)projects on a chosen topic.
Prerequisites. There are no specific prerequisites.
Assessment. In order to pass, students should attend the lectures and hand in a final report. The latter can be of manifold nature: from philosophical assessments of methods or phenomena, design of experiments to test model predictions on human behavior, to simulation results or mathematical characterizations of particular dynamics.
Participants. At most 15
Contact. {t.s.brochhagen, r.a.m.vanrooij}@uva.nl
Date, time & location | Topics | Slides | Reading |
---|---|---|---|
Monday, 16.01.2017 Time: 10:00-14:30 Room: F1.15 |
Basic game theoretic concepts, signaling games, conventions, population dynamics, ESS |
slides 1 slides 2 slides 3 slides 4 |
Benz, Anton, Gerhard Jäger, and Robert van Rooij (2005): An Introduction to Game Theory for Linguists Read before session! |
Tuesday, 17.01.2017 Time: 10:00-12:30 Room: F1.15 |
Game theoretic pragmatics | slides | Michael Franke & Gerhard Jäger (2014): Pragmatic Back-and-Forth Reasoning (suggested reading) |
Wednesday, 18.01.2017 Time: 10:00-12:30 Room: F1.15 |
The evolution of linguistic structure | slides | Huttegger, S. M., & Zollman, K. J. S. (2013): Methodology in Biological Game Theory (suggested reading) |
Thurday, 19.01.2017 Time: 10:00-12:30 Room: F1.15 |
Categorization, vagueness, stereotypes and sim-max games |
slides 1 slides 2 |
Gerhard Jäger and Robert van Rooij (2007): Language structure: psychological and social constraints (suggested reading) |
Monday, 23.01.2017 Time: 10:00-14:00 Room: F1.15 |
Meetings for project discussion |
◆ Costly signaling in evolution and/or pragmatics. For example, politeness, M-implicatures, honest signaling.
◆ The evolution of the a priori. For example, the emergence of causality, logic, or shared conceptual/meaning spaces
◆ Bounded rationality, mutual reasoning & linguistic choice. For example, an analysis on k-level reasoning (see V. Crawford's slides for a quick overview or Franke & Jaeger 2014 for a linguistic application of the idea) and its bearing on pragmatic inferences, e.g. "the rationality of bounded rationality", or a discussion of the consequences of heterogeneity in populations for the derivation of pragmatic inferences and its bearing on experimental results/linguistic theorizing (cf. Franke & Degen 2016).
◆ Lexical uncertainty. Bergen et al. (2016) recently proposed a generalization of models of rational language use by assuming that speakers take multiple lexica into consideration. This proposal is new, so little has been said about (i) its adequacy, in particular for manner implicatures but also more broadly, (ii) possible consequences for other standard results in this tradition, nor about (iii) other phenomena it may be applied to. The final report may, e.g., take the form of an answer to this article or apply this approach to other phenomena.
◆ Emergence of biases. For example, the prosecutors fallacy, the conjunction fallacy, or the evolution of stereotypes.
◆ Population-level/aggregated outcomes and their bearing on individual-level predictions. On the one hand, experiments in the iterated learning paradigm have been argued to, to some extent, reflect inductive learning biases of individuals. Similarly, results from experimental pragmatics are standardly aggregated but argued to inform us about (average?) individual-level choices in pragmatic reasoning. On the other, recent proposals have stressed the influence of the environment on population-level dynamics, e.g., its noisy perception, frequency effects and errors in production and comprehension without or with less influence of inductive biases, as well as the heterogeneity of individuals, e.g., in deriving pragmatic inferences. Notwithstanding, not much has been said about the theoretical and methodological implications of this apparent contrast.
◆ The evolution of ethics.
◆ Choice mechanisms and linguistic behavior. Linguistic choice is normally modeled as corresponding to 'rational' behavior, i.e., (expected) utility maximization. However, this needs not be so. There are a number of other mechanisms, e.g. regret minimization, that allow for a different subjective representation of a decision situation coupled with a method of choosing an act based on this representation. The project may, e.g., focus on the motivations for adopting a different choice mechanism and to work out what the consequences are for linguistic phenomena classically captured under utility maximization -- or those it fails to capture.