The symbol grounding problem asks how the symbols in a cognitive system acquire meaning. Forthirty-five years, the question has been framed as a problem of connection: how do internalrepresentations relate to the external world? This paper argues that the framing is the problem. Thegrounding literature has systematically studied intra-cranial mappings — relations betweensensorimotor and linguistic representations within the brain — while treating these as evidence forsymbol-world correspondence. But a simple observation undermines the correspondenceinterpretation: paradigmatically "grounded" human agents routinely and systematically err. Ifgrounding were symbol-world correspondence, grounded agents would not make the errors theymake. I propose, abductively, that grounding is a metacognitive feeling in the sense established byKoriat, Proust, and Arango-Muñoz: a heuristic monitoring signal, generated by the agent'sprediction-error-minimizing machinery, indicating that the agent's current predictive model isperforming within acceptable parameters. The feeling is adaptive but not veridical — it tracks thesmoothness of the predictive process, not correspondence with the world. This reframing replaces thequestion "how do symbols connect to the world?" with the question "how does an agent detect whenits predictive model fails?" — shifting the locus of the problem from ontology to error epistemology.I develop a three-layer architecture of epistemic adequacy: (1) the individual predictive model,which generates the grounding feeling; (2) the individual metacognitive loop, which detectsprediction failure; and (3) social error correction, which operates through consensus formation andgraded retrospective accountability. This architecture explains how societies achieve reliableknowledge without any individual having veridical world-access, and why systems lackingmetacognitive monitoring — such as current large language models — hallucinate as a structuralconsequence of their architecture rather than as a failure of world-contact.
Franny Philos Sophia (Tue,) studied this question.