Michael Franke (University of Tübingen) & Michael Henry Tessler (Stanford University)
For several decades, the study of language use and contextually mitigated meaning — the study of pragmatics — has been regarded as a wastebasket in which to dump unexplainable loose ends in service of more rigorous formal analyses. This view of pragmatics is no longer adequate. Owing to a dedicated shift towards experimental studies of pragmatic phenomena and the rise of suitable formal frameworks, a data-driven computational pragmatics is now possible. This course introduces ideas, experimental techniques and programming tools that feed recent efforts to model pragmatic inferences as probabilistic social cognition, in which interlocutors reason about each other’s choice of utterance and interpretation. We will introduce the Rational Speech Act model (Frank & Goodman 2012) and several of its extensions to cover such diverse phenomena as conversational implicatures, context-dependence, non-literal interpretations (hyperbole, metaphor, …), or politeness. We will discuss implementations of these models in WebPPL, a general-purpose probabilistic programming language. We demonstrate how computational pragmatic models can be evaluated against experimental data using Bayesian data analysis within WebPPL. In practical course units, participants will be able to explore model implementations and data analysis in an easily accessible browser-based implementation of WebPPL.
- Franke, M., & Jäger, G. (2016). Probabilistic pragmatics, or why Bayes’ rule is probably important for pragmatics. In Zeitschrift für Sprachwissenschaft 35(1):3–44.
- Goodman, N. D., & Frank, M. C. (2016). Pragmatic language interpretation as probabilistic inference. Trends in Cognitive Sciences.