Anatomical Critique of LLMorphism: Ontological Reductions and Epistemological Fault Lines in Human Cognition under Large Language Model Metaphors

Abstract

K SC This anatomical review examines Valerio Capraro's (2026) conceptualisation of LLMorphism—the biased belief that human cognition operates analogously to large language models (LLMs)—as introduced in arXiv:2605.05419. The paper is dissected sectionally: from its diagnosis of reverse inference arising from conversational fluency, through mechanisms of analogical transfer and metaphorical availability, to distinctions from adjacent constructs and projected societal impacts. Pros include the construct's novelty, careful differentiation, and timely intervention in public AI discourse; cons encompass its speculative character, absence of empirical validation, and under-specification of architectural dissimilarities between transformers and biological neural systems. At least ten competing models of mind and cognition—including predictive processing, embodied cognition, computationalism, enactivism, extended mind, dual-process theory, intentional stance, combinatorial creativity frameworks, argumentative theories of reasoning, and phenomenological accounts—are compared and contrasted. Ontological verticals (nature of cognition as embodied, agentic, and historically situated versus statistical pattern completion) and epistemological verticals (fluency as surrogate for justification; confabulation versus grounded inquiry) are analysed in depth. Also available on Academia.edu: https://www.academia.edu/attachments/134792343/download_file?s=portfolio