Synthetic Cognition: A Taxonomy for Cognitive Artifacts in the Era of Large Language Models

This essay proposes the formula "Synthetic Cognition" as a rigorous alternative to the expression "Artificial Intelligence" (AI). Through critical analysis, it argues that current computational artifacts achieve an isomorphism limited to functions (cognition), without possessing the necessary conditions for the attribution of mental states (intelligence). Positioning this thesis within the contemporary debate, the text engages with the objections of computational functionalism (Dennett), enactivism (Varela, Thompson), and recent theories on "distributed understanding" (Shanahan), analyzing the semantic limits of Large Language Models (LLMs). The proposed redefinition provides an empirical falsifiability criterion and an operational matrix for legislators (e.g., the AI Act), which is necessary to correctly attribute liability to human agents and ethically frame the technological infrastructure. Note: Manuscript are also available for download in 6 major languages (EN, IT, FR, DE, ES, SR) via the linked Zenodo repository.