Natasha Korotkova (Utrecht University)
Question bias, private beliefs and common knowledge

Various linguistic devices can be used to channel information about the speaker's beliefs, their communicative goals and the overall architecture of discourse. In this talk, I focus on one such device, or rather, a family of phenomena, referred to in semantic literature as "biased questions". Such questions differ from canonical inquiries for information in that they also convey the speaker’s attitude towards the truth or likelihood of the prejacent of the question (cf. English "Don't you like snowshoeing?") and have been typically viewed as means of conversation negotiation. In particular, they have been analyzed as updating discourse commitments of the interlocutors or as a way for the speaker to mediate, and possibly manipulate, the common ground. In this talk, I propose to shift from those discourse-based approaches to biased questions in favor of a purely doxastic view and offer a novel conceptualization of certain types of bias as a linguistic phenomenon sensitive to non-monotonic belief revision. This, in turn, opens up possibilities for connecting the phenomenon that has been studied almost exclusively within linguistics to a broader research agenda on belief formation and social reasoning.

Slides

Peter van Elswyk (Northwestern University)
Knowledge in the background

Conversation unfolds in a manner that is sensitive to what information is and isn't shared between participants. To be shared or not depends, in part, on whether each participant takes the same propositional attitude towards particular information. When information is shared, for example, this attitude is the attitude that we (take ourselves to) take towards the information, and represent others as taking towards it. But which attitude is this? The assumption in our workshop's name is that belief is the relevant attitude. In this talk, I will argue that knowledge is. To defend this conclusion, I will identify a number of properties that the relevant attitude must have, and argue that knowledge alone has these properties. My arguments will be empirical in nature, and draw on recent findings from the cognitive sciences regarding factive mindreading.

Dan Harris (Hunter College, CUNY)
Cognitive pluralism about common ground

Common ground is the information that the members of a group take themselves to share for the purposes of their interactions with each other. Linguists, philosophers of language, and psychologists have found lots of reasons to think that it does important work in language use and human communication. What I will be interested in is the following psychological question: what has to be going on in the group members’ minds in order for some information to be common ground for them, and by what process do they enter into that state? In this talk, I will defend a pluralistic answer to these questions, according to which there are a number of interestingly different ways to treat information as shared, and range of routes that we take to those states. My position is an example of cognitive pluralism, which is the view that minds typically have many different ways to solve their most important problems, each solution having its own costs and benefits that make it useful in different scenarios.

Poster session and coffee

Ramia Maria Koschel (University of Tübingen)
On the role of temporal knowledge in the construction of meaning

Marie Boscaro & Alda Mari (CNRS, Paris)
Accommodating background information in social media

Susanne Dietrich, Asya Achimova, Martin V. Butz & Bettina Rolke (University of Tübingen)
How does semantic causality influence the interpretation of statements in discourse?

Susanne Dietrich & Bettina Rolke (University of Tübingen)
Context-dependent understanding of non-verbal utterances in people with Asperger's syndrome

Ivan Rygaev (University of Tübingen)
Two common grounds are better than one

[Canceled] Henry Schiller (University of Sheffield)
Representing relevance

Jennifer Saul & Ray Drainville (University of Waterloo)
Can’t unsee: Perspectival dogwhistles and deniability (via Zoom)

Given the current state of the world it is not surprising that we have seen increasing academic attention to manipulative political speech, particularly that which promotes hatred and misinformation. One key mechanism that has seen a great deal of attention is the use of linguistic dogwhistles. In previous work, we expanded this discussion to include an exploration of visual dogwhistles. In this paper, we aim to build on this by exploring previously undiscussed forms of communication — in particular, AI-generated imagery using double-image effects and subliminal messages. We also examine related phenomena such as “Satanic” backwards messages in music, subliminal advertising, and some manipulative safety vests for horse riding. Whilst these examples have all or nearly all of the features that have been posited as necessary for dogwhistles, a close examination reveals that they lack one that has commonly been thought to follow from these: plausible deniability, sometimes known simply as ‘deniability.’ They also share another key feature with each other: the understanding of their hidden messages turns not on background beliefs, interests or practices, but rather on how they are perceived. They are a matter, quite literally, of perspective — in the sensory meaning of the term. Hence, we call these perspectival dogwhistles.

Slides

Elin McCready (Catalan Institution for Research and Advanced Studies)
Language, norms and identity

In this talk I argue that the perspectives agents hold — their ideologies, beliefs, and identities — can be extracted from their linguistic behavior. Specifically, because social meanings relate to speaker values and social beliefs, their language use tends to reflect these values. Taking such meanings and treating them in terms of self-ascription, together with norms of speech relating to how particular pieces of language are properly used, yields a `public perspective’ in partial correspondence to the notion of agential identity proposed in the recent philosophical literature.

Slides

Paula Rubio-Fernández (Max Planck Institute for Psycholinguistics)
The cognitive trinity of common ground

Human communication is built around interlocutors’ common ground (CG), or the information they assume to share. Despite having been the focus of intense interdisciplinary research for more than 60 years, we do not yet understand how CG works, or even what exactly it is. In this talk I will introduce a new research program that is essential to understanding CG: I propose to study CG as a product of cultural evolution. This approach requires identifying (i) those cognitive capacities that are required for the emergence of CG in human cognition, and (ii) how those capacities interact in (a) the development of CG through children’s social learning across cultures; (b) its formation through social interaction across the lifespan, and (c) its management in conversation across languages. I hypothesize that forming and using CG is a complex human ability that emerges from the interaction of three cognitive capacities — joint attention, shared memory, and the use of reference systems — under a rationality principle. This is what I informally call the Cognitive Trinity of Common Ground, which could also be described as a naïve model of rational memory.

With the commentary from Britta Stolterfoht (University of Tübingen).

Judith Tonhauser (University of Stuttgart) & Greg Scontras (University of California, Irvine)
Prior beliefs and projection: An RSA-based approach

Listeners sometimes infer the truth of content in the scope of entailment-canceling operators, for example inferring that Charley speaks Spanish from “Cole doesn’t know that Charley speaks Spanish”. These so-called projection inferences are sensitive to listeners’ subjective prior beliefs about the world (e.g., Mahler 2020, 2022; Degen & Tonhauser 2021). In this talk, we present an analysis, formalized as a computational cognitive model, of how prior beliefs modulate the projection of the content of the clausal complement of “know” in utterances of negated declaratives. Our model, formulated within the Bayesian Rational Speech Act framework (Frank & Goodman 2012), derives projection from the lexical entailments of “know” and sensitivity to the Question Under Discussion (QUD; as do Abrusán 2011; Simons et al 2017), as well as reasoning about utterance informativity relative to private speaker assumptions (Qing et al 2016; Warstadt 2022). Crucially, our model predicts projection for “know” without encoding the inference via a lexically-specified presupposition. Our model goes beyond existing analyses by also making predictions about the projection of the content of the clausal complement of nonfactive “think” in negated declaratives (“Cole don’t think that Charley speaks Spanish”), as well as about “know” and “think” in positive declaratives (“Cole knows/thinks that Charley speaks Spanish”) and positive and negative matrix sentences (“Charley speaks/doesn’t speak Spanish”). We evaluate our model’s predictions against new behavioral evidence that listeners’ prior beliefs about the content of the clausal complement (e.g., how likely it is that Charley speaks Spanish) modulate the strength of each of these types of inferences.

Slides

With the commentary from Todd Snider (University of Tübingen) (slides).

Asya Achimova, Michael Franke & Martin V. Butz (University of Tübingen)
Indirectness as a path to common ground management

Speakers may be uncertain as to whether certain beliefs are shared in conversation, and as a consequence, whether they belong to the common ground. In this talk, we argue that indirect utterances, such as “The election outcome was interesting!”, can be viewed as a tool for common ground management: They allow speakers to probe the beliefs of the listener without exposing their own beliefs publicly. We offer a Rational Speech Act model of indirect communication that captures (1) the speaker’s choice of indirect utterances; (2) the inferences conversation partners draw about the meaning of indirect utterances when they occur as a response in a conversation. We show that model predictions qualitatively match data from three behavioral experiments.

Slides

Mandy Simons (Carnegie Mellon University)
Whose beliefs? Speaker beliefs as background information

This talk starts from the observation of cases of successful communication in which the listener interprets an utterance in light of information which she does not accept. These cases demonstrate that information available to interlocutors for use in interpretation is neither practically, nor normatively, restricted to the common ground. Instead, available information must include information that the listener can infer to be assumed by the speaker. I spell out this idea using the recently proposed Multiple Perspectives Model (Heller & Brown-Schmidt 2024). This provides the basis for a revised understanding of the notion of presupposition as assumptions that the listener must attribute to the speaker in order to make sense of the speaker’s utterance. The notion of making sense will be given a precise characterization in Gricean terms. This view of presupposition, though, raises difficult questions: If presuppositions are merely assumptions of the speaker, why is it sometimes communicatively or socially problematic for a speaker to presuppose things that the listener rejects? Why does presupposing (by a speaker) sometimes seem to implicate the listener, in particular in cases of problematic presuppositions? In the talk I will attempt to make progress towards answering these questions.