A Pragmatic Account of the Weak Evidence Effect

Abstract

Language is not only used for neutral information; we often seek to persuade by arguing in favor of a particular view. Persuasion raises a number of challenges for classical accounts of belief updating, as information cannot be taken at face value. How should listeners account for a speaker’s “hidden agenda” when incorporating new information? Here, we extend recent probabilistic models of recursive social reasoning to allow for persuasive goals and show that our model provides a pragmatic account for why weakly favorable arguments may backfire, a phenomenon known as the weak evidence effect. Critically, this model predicts a systematic relationship between belief updates and expectations about the information source: weak evidence should only backfire when speakers are expected to act under persuasive goals and prefer the strongest evidence. We introduce a simple experimental paradigm called the Stick Contest to measure the extent to which the weak evidence effect depends on speaker expectations, and show that a pragmatic listener model accounts for the empirical data better than alternative models. Our findings suggest further avenues for rational models of social reasoning to illuminate classical decision-making phenomena.

Reconciling truthfulness and relevance as epistemic and decision-theoretic utility.

“…The important point is that the target cell under an epistemic QUD is determined solely in virtue of its relationship to some truthhood. Recent work outside the scope of our analysis has extended these techniques, allowing a speaker to be underinformative or deceptive in service of other conflicting non-epistemic goals, such as appearing competent (Yoon et al, 2018) or prosocial (Yoon et al, 2020), or being persuasive in service of a hidden agenda (Barnett et al, 2022).…”

“…Various epistemic goals may then be derived from differing preferences over real-world actions. Persuasion (Barnett et al, 2022;Mercier & Sperber, 2011) or deception (Oey et al, 2022) could emerge as short-term strategies to achieve speaker-serving actions, while truthfulness (Sbardolini, 2022) or politeness (P. Brown & Levinson, 1987) could signal aligned interests (Yoon et al, 2018(Yoon et al, , 2020 to preserve long-term cooperation (Baxter, 1984). 18 Finally, humans use language to enforce social norms (Li & Tomasello, 2021;Vaish et al, 2011) and commitments (Kanngiesser et al, 2017;Ostrom et al, 1992).…”

Reconciling truthfulness and relevance as epistemic and decision-theoretic utility

“…Self-serving epistemic goals may then emerge from differing preferences over real-world actions. In these circumstances, a speaker might attempt to appear polite (Yoon et al, 2020) or even deceive the listener (Barnett et al, 2022;Oey et al, 2022) in order to benefit from their future actions.…”

“…How, then, might we quantify such a "worthwhile difference"? Theorists have long potentially conflicting epistemic goals, such as appearing competent (Yoon et al, 2018) or prosocial (Yoon et al, 2020), or persuasion in service of a hidden agenda (Barnett et al, 2022).…”

The rejection game

“…Beyond degree-theoretic generalizations of the current framework there are probabilistic extensions, such as Bayesian pragmatics (see Franke & Jäger, 2016, and references therein), which includes models of communication under uncertainty or conflict of interest such as the Rational Speech Act model (Frank, 2017;Franke, 2011;Goodman & Frank, 2016;Sumers et al, 2021). This literature provides a further bridge between the two areas of study we mentioned, interpretation, and persuasion, by offering some insight into the cognitive science of social understanding (Barnett et al, 2022;Goodman & Stuhlmüller, 2013;Oey et al, 2023;Vignero, 2022). We regard it as a strength of a game-theoretic analysis of Stalnakerian conversations that speech act theory can then be seen as central to multiple areas of inquiry.…”