私善欲求与人工智能的道德主体性——基于生命完整性的责任归属判准

中文摘要

人工智能自主性上升，使“责任鸿沟”与道德主体性问题日益突出。既有讨论多以理性能力、感受能力或功能执行为判准，但这些判准仍不足以同时说明两个问题：何种存在能够把规范作为“与我有关”的行动理由，以及何种存在有资格成为最终的道德归责主体。本文基于作者《私善欲求论》的理论框架，提出“私善欲求”应被理解为完整道德主体成立的动力性必要条件。所谓私善欲求，是生命主体为了维持自身存在、改善自身状态并实现自身发展的内在滋养性动力结构，展开为自保、自顾、自富、自美、自乐、自怜、自思、自信、自强、自尊、自由十一项根欲求。本文所说的“生命完整性”并非把道德主体资格简单限定于特定生物材料，而是指自我维持、价值感受、持续身份与自我关切形成统一结构。私善欲求不是完整道德主体的充分条件；规范理解、行为控制与责任承担同样不可缺少。以此判准审视，当前人工智能虽具有强大的表征能力、规则执行能力和一定的道德决策模拟能力，但就现有系统的构造方式与可获得证据而言，尚无充分理由认为其具有内生需要、第一人称感受、持续自我价值关切及由此形成的真实私善欲求。因此，当前AI不宜被视为完整道德主体。AI造成损害时，最终道德责任原则上不应转移给AI本身，而应依据控制能力、因果贡献、收益关系、风险预见与注意义务等因素，在开发者、部署者、使用者及制度安排之间进行追溯和分配。本文同时回应婴儿、动物、法人、功能主义、生物主义与未来人工主体等可能反驳，并对未来人工主体的可能性保持条件性开放

Abstract

The rise of artificial intelligence autonomy makes the "responsibility gap" and the question of moral agency increasingly salient. Existing discussions mostly use rational capacity, feeling capacity, or functional execution as criteria, but these criteria remain insufficient to simultaneously explain two questions: what kind of being can take norms as action reasons that are "relevant to me," and what kind of being is qualified as the ultimate subject of moral accountability. Based on the theoretical framework of the author's book "On the Desire for Private Good," this paper proposes that the "desire for private good" should be understood as the dynamic necessary condition for the establishment of a full moral subject. The desire for private good, as used here, is the intrinsic nurturing motivational structure by which a living subject maintains its own existence, improves its own state, and realizes its own development, unfolding into eleven root desires: self-preservation, self-care, self-enrichment, self-beautification, self-joy, self-compassion, self-reflection, self-confidence, self-strengthening, self-esteem, and self-freedom. "Life completeness," as used here, does not simply restrict moral subject qualification to particular biological materials, but refers to the formation of a unified structure of self-maintenance, value feeling, persistent identity, and self-concern. The desire for private good is not a sufficient condition of the full moral subject; normative understanding, behavioral control, and the assumption of responsibility are equally indispensable. Judged by this criterion, although current AI possesses powerful representational capacities, rule-execution capacities, and a certain capacity to simulate moral decision-making, given the way existing systems are constructed and the available evidence, there is no sufficient reason to hold that it has endogenous needs, first-person feeling, persistent self-value concern, and the genuine desire for private good formed thereby. Current AI should therefore not be regarded as a full moral subject. When AI causes harm, the ultimate moral responsibility should in principle not be transferred to AI itself, but should be traced and distributed among developers, deployers, users, and institutional arrangements according to factors such as capacity for control, causal contribution, benefit relationships, foreseeability of risk, and duty of care. The paper also responds to possible objections from infants, animals, legal persons, functionalism, biologism, and future artificial subjects, while remaining conditionally open to the possibility of future artificial subjects.

关键词：私善欲求；生命完整性；人工智能；道德主体；责任鸿沟；第一人称体

Keywords: Desire for private good; life completeness; artificial intelligence; moral subject; responsibility gap; first-person experience