| 摘要: |
| 在司法人工智能明确定位于“辅助”后，未来中国司法人工智能的发展方向仍应是提升辅助型司法AI之“智”。借鉴“德雷福斯模式”，基于AI“能”与“智”二维可分离关系，我国可以建构辅助型司法AI“能—智分合”模式。当前提升司法AI“智”存在因果判断和价值判断的技术瓶颈，但并非没有技术解决方案。我国辅助型司法AI“智”的提升可通过三种技术方案分阶段实现： 一是“与法官一起思考”，以可解释性为抓手，在形式层面实现可用性的突破；二是“与法官的思考对齐”，在法律专业技能与职业伦理的约束下推进价值对齐；三是“像法官一样思考”，攀登因果判断和价值判断的阶梯，在“辅助”定位下增强对复杂裁量问题的推理支撑。 |
| 关键词: 司法人工智能 辅助型司法AI 德雷福斯模式 可解释性 价值对齐 |
| DOI： |
| 分类号: |
| 基金项目:教育部人文社会科学重点研究基地重大项目（2020—2024）“科技与人权的关系研究”（项目编号： 20JD820004） |
|
| “Capability Intelligence Separation and Integration”: A Phased Development Model for Judicial AI |
|
Sun Xiaoxia & Wei Yiming
|
|
|
| Abstract: |
| Even after judicial artificial intelligence has been explicitly positioned as “assistive”, the future trajectory of judicial AI in China should still aim to enhance the “intelligence” of assistive judicial AI. Building on the two-dimensionally separable relationship between AI's “capability” and “intelligence”, and drawing on the Dreyfus model, this article proposes a “capability–intelligence separation-and-integration” model for the development of China's assistive judicial AI. At present, further improvements in judicial AI's “intelligence” face technical bottlenecks in causal judgment and value judgment, but this does not mean that viable technical pathways are absent. The enhancement of “intelligence” in China's assistive judicial AI can be advanced in stages through three technical schemes. The first is “thinking alongside judges”, using explainability as a key lever to achieve a formal breakthrough in practical usability. The second is “aligning with judges' reasoning”, promoting value alignment under the constraints of legal expertise and professional ethics. The third is “thinking like judges”, ascending the ladder of causal and value judgment so as — within the assistive mandate — to strengthen inferential support for complex discretionary adjudication. |
| Key words: Judicial AI, Assistive Judicial AI, Dreyfus Model, Explainability, Value Alignment |
