Modeling Reference Production as the Probabilistic Combination of Multiple Perspectives

Abstract

While speakers have been shown to adapt to the knowledge state of their addressee in choosing referring expressions, they often also show some egocentric tendencies. The current paper aims to provide an explanation for this "mixed" behavior by presenting a model that derives such patterns from the probabilistic combination of both the speaker's and the addressee's perspectives. To test our model, we conducted a language production experiment, in which participants had to refer to objects in a context that also included a visually misleading object (e.g., a crayon shaped like a Lego brick) whose function was either known to both partners, or known just to the speaker but not the addressee. Modeling results indicate that the experimental findings cannot be explained by assuming that speakers tailor a referring expression solely to their own perspective or to the perspective of their addressee. Instead, accounting for the behavioral pattern requires an approach where both perspectives influence the choice of referring expressions. Nevertheless, in our situation, speakers consider their partner's perspective more than their own.

Taking common ground into account: Specifying the role of the mentalizing network in communicative language production

“…Speakers produced more expected responses in the communicative linguistic control condition compared to the communicative privileged ground condition, t(19) = 3.07, p < 0.01, and in the communicative visual control condition compared to the communicative privileged ground condition, t(19) = 3.68, p < 0.01. These findings are in line with previous studies with similar designs that showed that speakers generally take their addressee’s perspective into account in communicative tasks, although they cannot completely ignore privileged ground information [ 9 – 13 ]. The results of the linguistic control condition are below ceiling, because speakers sometimes produced overinformative responses by including a modifier when it was not necessary (see also the results of the advisable linguistic control condition in [ 10 ]).…”

“…Speakers generally took their addressee’s perspective into account when designing referring expressions in the communicative blocks, although they failed to ignore privileged information on all trials. These results are in line with previous findings using similar paradigms [ 9 – 13 ] and mostly results replicate the findings of our previous eye-tracking study [ 10 ]. Speakers produced slightly fewer expected responses in the communicative conditions in the current study compared to the previous study.…”

“…Previous psycholinguistic research has shown that speakers are generally able to distinguish between common and privileged ground information in their utterances, but that they cannot completely ignore privileged ground information (e.g., [ 9 – 13 ]). For example, in a previous eye-tracking study using the same paradigm as the communicative blocks in the current study [ 10 ], we found that speakers generally take their addressee’s perspective into account in privileged ground trials, but sometimes produce descriptions from their own perspective.…”

The production and comprehension of referring expressions: Definite description

“…Our investigation of the linguistic context has taught us what information needs to be represented in the discourse model, which, in addition to linguistic information, must include (a) as-yetunmentioned properties of perceptually available objects and their relationships and (b) the identity of interlocutors for each conversational exchange. Finally, our investigation of referential domains revealed that a particular definite description is not always interpreted relative to a single, categorical referential domain, but is better captured when combining the influence of two domains as originally proposed by Heller et al, (2016) and Mozuraitis et al, (2018).…”

“…It is our hope that insights from the investigation of definite descriptions would prove useful in guiding the study of demonstrative descriptions. F I G U R E 7 An illustration of the experimental setup in Mozuraitis et al (2018). The speaker and the listener could see the same two objects: the typical crayon and the crayon shaped to look like a Lego.…”