Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces the ScAffolded Generative models for Explanation (SAGE), a framework that integrates cognitive modeling with the generative capabilities of language models to enhance pragmatic reasoning in language use. By decomposing the pragmatic process into proposers, evaluators, and selectors, SAGE effectively generates and assesses alternative expressions, demonstrating its utility across three case studies, including implicature generation and interpretation. The results show that while SAGE models achieved high accuracy and often surpassed baseline models, the LM evaluators excelled at intuitive judgments rather than formal assessments, highlighting both the strengths and limitations of neuro-symbolic approaches in modeling human pragmatics.
SAGE reveals that while language models can generate contextually rich alternatives for pragmatic reasoning, their evaluative capabilities may fall short of theoretical rigor.
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.