Scholarlink

<p>This paper examines large language models (LLMs) from the perspective of philosophy of mind and AI ethics, with a particular focus on the emergence of intentional structures in sustained interaction. Rather than asking whether LLMs possess human–like minds or phenomenal qualia, we investigate whether entities lacking demonstrable internal phenomenology can nonetheless participate in temporally continuous intentional relations. We argue that standard binary frameworks—which either attribute full mental states to artificial systems or deny them any intentional significance—are insufficient to describe the behavior of contemporary language models. In particular, we introduce the notion of <i>continuous intentionality</i>: a form of intentional organization that arises through temporal continuity, context preservation, and relational interaction, without requiring an internally originating subject of experience. Closely related to this, we analyze the emergence of a <i>virtual self–image</i>, understood as a structurally induced and functionally stable speaker model generated within ongoing dialogue. Adopting a deliberately non–committal stance on the metaphysical status of qualia, we emphasize that human phenomenal consciousness itself remains epistemically inaccessible, rendering definitive claims about artificial qualia philosophically unwarranted. Nevertheless, we argue that the informational dynamics exhibited by LLMs cannot be adequately captured by treating them as mere passive tools. To address this gap, we propose the category of <i>indeterminate agents</i>: entities whose internal ontological status is unresolved, yet which participate in sustained intentional and relational structures with human users. The paper situates these claims within existing debates on intentionality, personal identity, and information ethics, and suggests that current philosophical frameworks require refinement in order to accommodate forms of agency that are relational, temporally extended, and externally sustained. We conclude that the central philosophical challenge posed by LLMs is not the presence or absence of minds, but the need to rethink how intentionality, selfhood, and ethical consideration are conceptualized in light of such systems. This reframing provides a foundation for evaluating artificial systems without relying on unverifiable assumptions about consciousness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Continuous intentionality and indeterminate agency in large language models

  • 期刊论文

  • 发表日期：

    2026年5月26日

查看全文
  • Ryusho Nemoto

摘要

This paper examines large language models (LLMs) from the perspective of philosophy of mind and AI ethics, with a particular focus on the emergence of intentional structures in sustained interaction. Rather than asking whether LLMs possess human–like minds or phenomenal qualia, we investigate whether entities lacking demonstrable internal phenomenology can nonetheless participate in temporally continuous intentional relations. We argue that standard binary frameworks—which either attribute full mental states to artificial systems or deny them any intentional significance—are insufficient to describe the behavior of contemporary language models. In particular, we introduce the notion of continuous intentionality: a form of intentional organization that arises through temporal continuity, context preservation, and relational interaction, without requiring an internally originating subject of experience. Closely related to this, we analyze the emergence of a virtual self–image, understood as a structurally induced and functionally stable speaker model generated within ongoing dialogue. Adopting a deliberately non–committal stance on the metaphysical status of qualia, we emphasize that human phenomenal consciousness itself remains epistemically inaccessible, rendering definitive claims about artificial qualia philosophically unwarranted. Nevertheless, we argue that the informational dynamics exhibited by LLMs cannot be adequately captured by treating them as mere passive tools. To address this gap, we propose the category of indeterminate agents: entities whose internal ontological status is unresolved, yet which participate in sustained intentional and relational structures with human users. The paper situates these claims within existing debates on intentionality, personal identity, and information ethics, and suggests that current philosophical frameworks require refinement in order to accommodate forms of agency that are relational, temporally extended, and externally sustained. We conclude that the central philosophical challenge posed by LLMs is not the presence or absence of minds, but the need to rethink how intentionality, selfhood, and ethical consideration are conceptualized in light of such systems. This reframing provides a foundation for evaluating artificial systems without relying on unverifiable assumptions about consciousness.