The Pragmatics of Understanding Sexual Innuendo: A Conditional Comparative Model of Human and AI Processing of Sexual Implicature — From the Perspective of Embodied Cognition and the Symbol Grounding Problem —

This paper reframes the understanding of sexual implicature in utterances—colloquially, “sexual innuendo”—not as a problem of lexical classification but as a pragmatic phenomenon at the intersection of speaker–hearer relationship, situational norms, embodied and affective response, and interactional history, and offers a comparative model for analyzing, under specified conditions, the performance gap between humans and current text-centric large language models (LLMs). Rather than mathematically “proving” this gap, the paper presents a testable model built from three operationally defined variables—context-dependency (C), embodiment-relevance (E), and normative variability (T)—together with task-specific expected-utility and loss functions, and outlines predictions and an experimental design for empirical research. The symbol grounding problem, embodied cognition, and pragmatics/relevance theory are examined together with their contested counter-positions, leading not to the general claim that “AI does not understand meaning,” but to the narrower claim that “in high-context, high-normative-variability tasks, current text-centric AI may exhibit information conditions and failure modes that differ from those of humans.” The aim of this paper is not to declare a standing human superiority but to provide an analytic framework useful for AI design, safety, education, and institutional policy.