Contextualising (Im)plausible Events Triggers Figurative Language

2026-04-09

AI Summary

This study investigates the relationship between literalness and plausibility in English subject–verb–object events, as well as the cognitive differences between humans and large language models (LLMs) in interpreting non-literal and implausible events. By constructing systematic triplets of plausible and implausible events and distinguishing between abstract and concrete semantic components, the research integrates human judgments with LLM-generated analyses. Findings reveal that humans adeptly differentiate and contextually interpret both types of anomalous events, whereas LLMs exhibit only superficial contextualization, frequently misinterpreting implausible events as acceptable non-literal expressions. These results indicate a systematic bias in current LLMs when processing semantic anomalies and underscore their limited capacity for human-like contextual reasoning, particularly in understanding figurative language.

Abstract

This work explores the connection between (non-)literalness and plausibility at the example of subject-verb-object events in English. We design a systematic setup of plausible and implausible event triples in combination with abstract and concrete constituent categories. Our analysis of human and LLM-generated judgments and example contexts reveals substantial differences between assessments of plausibility. While humans excel at nuanced detection and contextualization of (non-)literal vs. implausible events, LLM results reveal only shallow contextualization patterns with a bias to trade implausibility for non-literal, plausible interpretations.

Problem

• plausibility
• literalness
• figurative language
• event triples
• contextualization

Innovation

• figurative language
• plausibility
• literalness
• large language models
• contextualization