A taxonomy of errors in biological and synthetic minds.

This paper presents a unifying framework for understanding cognitive errors in both artificial and biological minds. It introduces three contributions. First, an eight-step model of the collective cognition process that distinguishes between reality, logos, data, information, meaning, understanding, insight, and knowledge. Second, a four-layer model of cognition: data, meaning, understanding, knowledge. We argue that most current AI systems operate between the data and information layers, and that most humans operate between information and insight. Second, a taxonomy of cognitive errors that classifies AI failures into three distinct categories: hallucinations (errors made to bridge knowledge gaps), hallucinOtions (errors of understanding), and glitches (system errors). We argue that hallucinotions (errors in which a system does not realize it lacks understanding) are not unique to AI. Instead, they are pervasive in human cognition as well. This claim, in relation to AI, is supported by recent empirical research on “potemkin understanding” in large language models (Mancoridis et al., 2025), which independently validates the distinction between factual errors and conceptual incoherence (first proposed by the author in October 2024). We also present empirical proof of conceptual understanding in human minds and in other biological minds. The framework has implications for AI development, epistemology, cognitive science, semiotics, education, and pedagogy.