代价的肉身——为什么人工智能无法跨越“责任鸿沟”

• 作者： 樊明迪 ¹
• 作者单位：
  1.  苏州大学
• 发布时间：2026-05-01

摘要: 现有的“责任鸿沟”（ResponsibilityGap）理论主要从认识论视角讨论人工智能行为的责任归属问题，却忽略了一个更为根本的本体论层面——行为主体是否必须为其决策付出“生存性代价”。本文提出，责任的本质并非纯粹的计算或推理结果，而是由于行为主体必须承担不可逆的物理代价而被“迫认”的。本文通过界定“生存性代价”概念，将其分析为热力学不可逆代价与必死性的双重结构，并借助存在主义哲学与身体现象学的资源，论证人类决策所伴随的神经生化改写、寿命熵增等不可逆代价，构成了责任归属的本体论基础。相较之下，人工智能的推理过程遵循热力学可逆计算原则，其“决策”不涉及任何生命性资源的耗散。

Abstract: This paper re-examines the "responsibility gap" in artificial intelligence ethics from an ontological perspective rather than the conventional epistemological approach. We argue that the foundation of responsibility lies in "existential cost"—the irreversible physical price that embodied agents must pay for their decisions and actions. This existential cost manifests as a dual structure: thermodynamic irreversibility rooted in the physics of information processing, and mortality embedded in the finitude of human existence. Drawing on existentialist philosophy (Heidegger's Being-toward-death, Sartre's "condemned to be free") and phenomenology of embodiment (Merleau-Ponty's account of the lived body), we demonstrate that human decision-making is necessarily entangled with irreversible neurochemical rewiring, physiological resource depletion, and temporal finitude. In stark contrast, AI systems operate through thermodynamically reversible computation—they can be paused, copied, rolled back, and do not consume life-based resources. This ontological asymmetry constitutes the unbridgeable responsibility gap. Our analysis offers both theoretical resources for understanding the nature of responsibility and normative guidance for human oversight of high-stakes AI decisions.

• 来自： 樊明迪
• 分类： 哲学 >> 伦理学
• 稿件状态： 未投稿
• 引用： PSSXiv:202605.00073 (或此版本 PSSXiv:202605.00073V1)
  DOI:10.12451/202605.00073
  CSTR:32012.36.PSSXiv.202605.00073
• 推荐引用方式： 樊明迪.代价的肉身——为什么人工智能无法跨越“责任鸿沟”.哲学社会科学预印本平台:https://zsyyb.cn/abs/202605.00073.[PSSXiv:202605.00073V1] (点此复制)