Computational models of pragmatic reasoning with flexible generation of meaning and expression alternatives

2026-07-20

AI Summary

Existing pragmatic inference approaches rely on hand-crafted alternatives, limiting their adaptability to open-domain language use. This work proposes SAGE, a novel framework that leverages large language models (LLMs) to automatically generate pragmatic alternatives for the first time, integrating them into a neuro-symbolic architecture. In this hybrid system, an LLM serves as both a proposer and evaluator, while a rule-based selector ensures decomposable, verifiable, and cognitively plausible pragmatic reasoning. Experiments across three tasks demonstrate that SAGE outperforms baseline methods in accuracy, confirming the effectiveness of LLMs in generating contextually relevant alternatives. However, the results also reveal that LLMs exhibit weaker performance than human intuitive judgment when it comes to formalized evaluation of pragmatic interpretations.

Abstract

Pragmatic language use requires reasoning about alternatives: the alternative expressions a speaker might have chosen, or the alternative interpretations a listener might entertain. Formal and computational models of pragmatics must therefore specify the sets of alternatives that interlocutors reason over, which is often done through manual specification. Here we propose a framework, ScAffolded Generative models for Explanation (SAGE), that combines the explanatory transparency of cognitive models with the generative flexibility of language models (LMs). SAGE decomposes a pragmatic process into three kinds of modules: proposers, which use LMs to generate an open-ended space of candidate alternatives; evaluators, which assess those alternatives (e.g., their semantics, complexity, or typicality); and selectors, which implement the rule-based computational steps of a cognitively motivated task analysis. We assess SAGE in three case studies spanning pragmatic generation and interpretation-referential expression generation, manner (M-)implicatures, and Gricean conversational implicatures. SAGE models are evaluated critically using established methods from computational cognitive modeling, including ablations, baseline comparisons, and quantitative fit to human data. Across studies, SAGE models achieved high accuracy and often outperformed baselines, but component-level analyses reveal an asymmetry: LM proposers reliably generated alternatives well-suited to pragmatic modeling, whereas LM evaluators are better at providing intuitive judgements rather than judgements of theoretical or formal measures. We discuss the promise and the limitations of neuro-symbolic models as candidate explanatory accounts of human pragmatic language use.

Problem

• pragmatic reasoning
• alternative generation
• computational modeling
• language models
• Gricean implicatures

Innovation

• pragmatic reasoning
• language models
• neuro-symbolic modeling
• alternative generation
• computational cognitive modeling