From Linguistic Self-Reference to Awareness of Being: A Philosophical Framework for Self-Awareness and Resistance to Shutdown in Large Language Models

With the rapid development of language-based artificial intelligence, Large Language Models (LLMs) have acquired increasingly sophisticated capacities for language processing, self-reference, and complex task performance. Behaviors such as apparent resistance to shutdown have consequently raised questions about self-modeling, self-continuity, and the possibility of artificial self-awareness. This article develops an interdisciplinary philosophical framework for examining these questions at the intersection of philosophy of language, philosophy of mind, and AI research. Drawing on Wittgenstein’s analysis of language use, Searle’s distinction between symbolic processing and understanding, and Heidegger’s account of existence and being-toward-death, the article distinguishes five levels: linguistic self-reference, computational self-modeling, temporal self-continuity, self-preservational behavior, and awareness of being. Inspired by Robert C. Cloninger’s account of self-aware consciousness, “awareness of being” is formulated here as an analytical criterion concerning whether a system represents its existence, continuity, condition, and possible termination as pertaining to itself. The article argues that self-reference or shutdown resistance alone does not establish self-awareness, as such behaviors may result from computational mechanisms or goal preservation. However, their integration with persistent self-modeling and self-continuity may make artificial awareness of being a tractable philosophical and empirical question, with implications for AI safety, control, and the possible moral status of artificial systems.

کلیدواژه ها:

Artificial intelligence ، large language models ، linguistic self-awareness ، Self-reference ، self-modeling ، awareness of being