Periodic Reporting for period 1 - PlauDe (Judging Plausible Deniability)
Période du rapport: 2022-09-01 au 2024-08-31
Plausible deniability (PD) is a linguistic strategy that consists in intentionally conveying content in such a way that the intention can later be denied. PD is essential in the success of e.g. covert threats (“It would be a shame if something were to happen to you”), indirect insults (“I admire your courage for wearing that dress”), or sexual advances (“Would you like to come up for a coffee?”). A rigorous, scientific account of PD therefore can shine light on cases of language use in uncooperative of otherwise socially complex scenarios. A scientific account of PD as a communicative strategy is most naturally coached in the framework of Gricean pragmatics, where communication is reduced to a special case of interaction between rational agents. The current cutting edge of this reductive approach uses a combination of computational cognitive modelling, most importantly the Rational Speech Act (RSA) framework, and experiments. However, PD presents several methodological challenges for current techniques. First, in line with traditional Gricean pragmatics current models assume that interlocutors by and large share their goals (Grice's cooperative principle), while PD is often motivated by a goal misalignment. Second, current models of language use do not model social norms, which play a crucial role in what makes PD a popular strategy. Third, the current explanatory models of pragmatics require a manual specification of all the features of the context which are open-ended, e.g. the possible utterances and states. In this context, PlauDe has two fundamental concurrent objectives. The first is to develop a unified scientific account of PD as a rational communicative strategy that can subsume the disparate variety of cases of PD. The second is to enrich current models of pragmatics with the mechanisms that are needed to make sense of PD, and to open-end them. PlauDe has three main pathways to scientific impact. First, techniques to open-end current explanatory models of pragmatics can be applied to phenomena other than PD, and can therefore have an impact in the field of computational pragmatics. Second, research on PD fits within a recent wave of interest in socially sophisticated language within the field of pragmatics. Third, understanding PD can be a springboard to better models of the interaction between language use and social norms, extending computational pragmatic beyond its traditional Gricean assumptions.
The work performed in the first year of the project can be divided into four parts. First, we have continued developing a model of argumentative language use. This first project requires some more work, mainly an experiment, before reaching the writing stage. Second, we have started developing a theoretically coherent picture of PD that is couched in a general picture of instrumental rationality. This second project represents substantial progress over the state of understanding of PD at when the grant started. Third, we have developed a model of politeness that starts encoding some of the ideas in the theory of PD. This model was presented at the 2023 CogSci conference. Fourth, we have started developing a new approach to modelling pragmatic language use in an open-ended way, which can be then applied to PD a special case. This project is now at the writing stage and we plan to submit it as a paper journal before the end of 2023.
The grant had to be interrupted prematurely after one year because the beneficiary secured a permanent position at the University of Amsterdam. The main results of the project at the current stage can be summarized as follows. First, we developed models of argumentative and polite language use, and supported them with human behavioural data. Second, we developed a novel computational framework for developing open-ended cognitive models of pragmatics, which can form the basis for a variety of models. Third, we started developing a theoretical account of PD that reduces PD to general principles of rational behaviour on the background of a system of social norms of accusation, exculpation, and punishment. This represents the state-of-the-art in scientific understanding of PD and would most naturally be written as a theoretical paper, but still requires further refinements before reaching the writing stage.