Search

Probabilistic models of language understanding are valuable tools for investigating human language use. However, they need to be hand-designed for a particular domain. In contrast, large language models (LLMs) are trained on text that spans a wide array of domains, but they lack the structure and interpretability of probabilistic models. In this paper, we use chain-of-thought prompts to introduce structures from probabilistic models into LLMs. We explore this approach in the case of metaphor understanding. Our chain-of-thought prompts lead language models to infer latent variables and reason about their relationships in order to choose appropriate paraphrases for metaphors. The latent variables and relationships chosen are informed by theories of metaphor understanding from cognitive psychology. We apply these prompts to the two largest versions of GPT-3 and show that they can improve performance in a paraphrase selection task., 7 pages, 1 figure

Included in this OSF Project are stimuli, Qualtrics surveys, data, and analysis scripts for a paper on how vertical eyes affects perceptions of emotional valence in faces. The paper associated with this project has been published in the Quarterly Journal of Experimental Psychology.

Initial round of data collection using this new paradigm -- manipulating Item Format (pictures or words), Label Length (short or long), as well as number of items (8, 10, 12, 15, 18).

Additional file 1. Supplementary materials.

Myriad linguistic cues, subtle and otherwise, have been shown to shape thinking and reasoning. Such linguistic framing effects are attributed variously to irrational biases, affective processes, and analogy. Across two studies, we explored pragmatic reasoning (PR) as a potential unifying mechanism underlying disparate framing effects. In Study 1, we adapted a measure of PR used in attribute framing research, assessing the ability to infer information subtly communicated by a speaker’s choice of frame. Our measure captured individual differences in PR distinct from cognitive reflection and mentalizing ability. In Study 2, PR predicted the effect of identity labels on attitudes towards illegal immigration, but not other framing effects. Two such effects were driven, however, by participants’ explicit recognition of the influence of the framing language. These results suggest that framing effects may indeed be mediated by PR, but that our measure of PR does not fully capture its mediating role.

Metaphors can both shape and reflect how we think about com- plex issues. Here, we explore the prevalence of WAR language in a large corpus (N = 1.63 million) of Tweets. We compare how different groups of people use the language of WAR to talk about different topics. In Study 1, we find that about 5% of Tweets about Covid-19 from the general public include WAR language, replicating prior work (Wicke & Bolognesi, 2020). In Studies 2 and 3, we find that politicians use WAR language much more often, while Reporters and Medical Experts use WAR language less often. The findings are relevant to current debates about the role of language in democracy, and to theo- ries of metaphor in communication.