I know what you're probably going to say: Listener adaptation to variable use of uncertainty expressions

Abstract

<p>Pragmatic theories of utterance interpretation share the assumption that listeners reason about alternative utterances that a speaker could have produced, but didn't. For such reasoning to be successful, listeners must have precise expectations about a speaker's production choices. This is at odds with the considerable variability across speakers that exists at all levels of linguistic representation. This tension can be reconciled by listeners adapting to the statistics of individual speakers. While linguistic adaptation is increasingly widely attested, semantic/pragmatic adaptation is underexplored. Moreover, what kind of representations listeners update during semantic/pragmatic adaptation -- estimates of the speaker's lexicon, or estimates of the speaker's utterance preferences -- remains poorly understood. In this work, we investigate semantic/pragmatic adaptation in the domain of uncertainty expressions like might and probably. In a series of web-based experiments, we find 1) that listeners vary in their expectations about a generic speaker's use of uncertainty expressions; 2) that listeners rapidly update their expectations about the use of uncertainty expressions after brief exposure to a speaker with a specific usage of uncertainty expressions; and 3) that listeners' interpretations of uncertainty expressions change after being exposed to a specific speaker. We present a novel computational model of semantic/pragmatic adaptation based on Bayesian belief updating and show, through a series of model comparisons, that semantic/pragmatic adaptation is best captured by listeners updating their beliefs both about the speaker's lexicon and their utterance preferences. This work has implications for both semantic theories of uncertainty expressions and psycholinguistic theories of adaptation: it highlights the need for dynamic semantic representations and suggests that listeners integrate their general linguistic knowledge with speaker-specific experiences to arrive at more precise interpretations.</p>

Semantic Adaptation to the Interpretation of Gradable Adjectives via Active Linguistic Interaction

“…This reveals they can adapt their representations to align them to those by another speaker. Similar evidence was reported by S. Heim et al (2020) for the vague quantifiers few and many, while Schuster and Degen (2019) reported a comparable adaptation mechanism for the expressions of uncertainty might and probably. Recently, Schuster and Degen (2020) further showed that this adaptation is best captured by speakers who update beliefs both about the interlocutor's lexicon and their utterance preferences.…”

“…We experiment with two conditions: one in which speakers are given the possibility to ask clarification questions to the target speaker before assessing the trial (Q); one in which they can only learn by passively interacting with it (C). This latter setting is loosely comparable to that by Yildirim et al (2016), S. Heim et al (2020), and Schuster and Degen (2019).…”

“…More in general, Solt (2018) proposed that the observed variability in the use of gradable adjectives would result from the speaker-and context-dependent weighting of the multiple dimensions that underlie many gradable adjectives. As a consequence, various speakers may have a different interpretation of what count as tall, small, or heavy in a given context, similarly to what has been reported for quantifiers (S. Heim et al, 2020;Yildirim et al, 2016) and expressions of uncertainty (Schuster & Degen, 2019). All these aspects-context-dependence, vagueness, and speaker variability-make gradable adjectives an interesting benchmark for testing semantic adaptation.…”

Strategic Use of (Un)certainty Expressions

“…For example, there is a considerable amount of inter-speaker variability (e.g., Wallsten et al, 1986) that might not be explainable with varying event probabilities. In a recent study, Schuster and Degen (2020) suggest that hearers use the speaker's identity when interpreting utterances: what expressions does a speaker preferably use and what thresholds do they associate with specific expressions? Schuster and Degen (2020)'s study tests how hearers adapt to speaker-specific thresholds when interpreting uncertainty expressions.…”

“…Thus, speakers aim to reduce uncertainty for the hearer by choosing the most informative utterance while also having personal preferences as to which utterance to choose. As was pointed out by Schuster and Degen (2020), while these considerations successfully capture inter-speaker variability, it is not clear how such an approach would capture the effect of additional communicative goals speakers may have when using uncertainty expressions. For example, a speaker may use 'might' to make a statement more moderate, or to be polite-goals which may vary by context even for the same speaker and which conflict with the goal of presenting the strongest possible information, in order to reduce uncertainty for the hearer.…”