Sariyar, MuratMuratSariyar2026-01-092026-01-092025-06-26https://doi.org/10.24451/arbor.1268410.3233/SHTI250690https://arbor.bfh.ch/handle/arbor/4631740588898As large language models (LLMs) like GPT-4 are increasingly deployed in clinical and administrative healthcare settings, questions about their conceptual grounding take on renewed urgency. While concerns about the lack of sensorimotor experience in symbolic AI systems have been long discussed in cognitive science and philosophy of mind, their practical implications in medicine remain underexplored. This paper revisits the grounding problem through the lens of contemporary healthcare applications, arguing that the unique demands of medical reasoning - interpretive nuance, ethical sensitivity, and contextual depth-amplify the limitations of ungrounded AI. By reframing classic debates, such as Searle's Chinese Room and the Symbol Grounding Problem, within real-world clinical contexts, we highlight specific risks that emerge when LLMs are treated as epistemic agents rather than tools.enArtificial intelligence (AI)Large Language Model (LLMs)artificial general intelligence (AGI)symbol groundingarticle