The Observer’s Blind Spot: Self, Other, and the Dilemma of Understanding in Artificial Intelligence

Declarative sentences (“S is P”) appear grammatically as objective statements, yet in pragmatic use they always imply an unarticulated speaker’s perspective. This structure of the self—simultaneously present and absent—constitutes a fundamental feature of human understanding and raises deep philosophical questions for AI language processing. Starting from the phenomenological thesis of the “non—objectifiability of the operating subject” (Husserl, Merleau‑Ponty), this paper analyzes the non‑closability of perspectival recursion, and examines the dual role of the “I” in language, the experience of selflessness in flow, and the asymmetry between self and other in understanding. On this basis, it proposes a functional solution: positioning AI’s understanding as “functional intersubjectivity”—that is, continuously managing perspective‑related uncertainty through interaction, rather than pursuing a complete representation of the self. This approach responds to critiques by Dreyfus and others regarding the possibility of AI understanding, while providing a phenomenological foundation for computational semantics.