摘要： 站在第四次工业革命的十字路口，人工智能、大数据和云计算等技术正带来一场司法文明的范式变革。一方面，新技术通过提升司法的智能化水平，增强了司法在社会中的治理效能，有利于满足公众对司法功能的期待。另一方面，新技术的介入也使得司法运作的逻辑和流程发生了改变，并带来系列法律、技术、伦理以及司法功能异化的问题，尤其是在深层次对司法领域的“人机关系”形成冲击。面对司法智能化转型的大趋势，我们需要在价值层面坚守法治的底线，防止技术理性侵蚀人文关怀的根基，共同探索数字正义的实现路径。有鉴于此，《探索与争鸣》编辑部与上海对外经贸大学法学院共同举办圆桌会议，邀请相关领域专家围绕数字司法图景、智能化实践探索、司法价值坚守等核心议题进行深入讨论。 傅郁林教授指出，AI法官是人工智能技术运用于司法活动的终极状态，而当下AI的发展程度距离担当起核心审判职能所需的法律专业化水平还有较大差距，但其在技术上仍具有可行性。而若要在制度上赋予AI法官以审判主体权限，由于司法所蕴含的人性特质与AI法官所欲实现的替代功能存在冲突，所以很难实现。王福华教授认为，数字技术给人类带来便捷和效率的同时，也衍生出诸多“数字不正义”现象。要通过纠正数字不公、修复数字侵害、制衡技术权力、实现程序正义和促进数字包容等途径，实现数字的矫正正义。唐力教授强调，人工智能的司法应用可能侵蚀人类作为司法主体的“主体性”与“创造性”，需要在人工智能司法应用的伦理准则、“人机关系”定位、应用范围，以及技术安全保障等维度采取具体措施，明确人工智能司法应用的限度。许多奇教授以智能化赋能金融案件“四大检察”为分析场景，认为技术层面的数据壁垒、算法局限，机制层面的协同不畅、权责不清，以及伦理和规范层面的价值冲突与法律滞后等问题，构成制约智能化赋能检察的核心。需要在破除数据壁垒、精准适配算法、筑牢安全机制等核心举措的基础上，同步消除金融案件办理过程中跨域协作、权责划分、规范适用等制度性障碍。张兴美教授以民事司法为例，指出当下的人工智能正作为智能的“第四方”在纠纷解决层面发挥了实质性作用，但也应警惕技术至上主义对民事司法的异化。应该秉持司法人类主体性理念、注重法官参与、坚持司法公开、加强系统监督，推动人工智能时代司法的公平正义。黄泽敏副教授构建了司法领域智能科技的介入的强式、中等和弱式三种标准。其中，中等和弱式应让位于强式标准，强式标准强调凡涉及法律判断的，智能科技不可介入。强式标准之所以成立，是因为除了形式要素之外，法律判断还复合了价值、道德、情感和权力等要素，而人工智能不能也不应该介入这些要素。季平平研究员指出，人工智能技术的发展为裁判文书的结构优化提供了可能路径。然而，由于法律推理内含事实认定与价值判断的双重属性，人工智能仍难以胜任具有规范创设与理由展开功能的文书自动生成任务。人工智能的合理定位，应是技术辅助工具而非替代者。在促进裁判文书结构变革的同时，应尊重法官对法律解释的最终权限。

Abstract: Standing at the crossroads of the fourth industrial revolution, technologies such as artificial intelligence, big data, and cloud computing are bringing about a paradigm shift in judicial civilization. Faced with the trend of judicial intelligence transformation, we need to adhere to the bottom line of the rule of law at the value level, prevent technological rationality from eroding the foundation of humanistic care, and jointly explore the path to realizing digital justice. In view of this, the editorial department of “Exploration and Free Views” and the Law School of Shanghai University of International Business and Economics jointly held a roundtable meeting, inviting experts in relevant fields to have in-depth discussions on core issues such as the digital judicial landscape, intelligent practice exploration, and adherence to judicial values.   Professor Fu Yulin pointed out that AI judges are the ultimate state of applying artificial intelligence technology to judicial activities, and the current level of AI development is still far from the level of legal specialization required to undertake core judicial functions, but it still has technical feasibility in theory. Professor Wang Fuhua believes that digital justice should be achieved through correcting digital injustice, repairing digital infringement, balancing technological power, realizing procedural justice, and promoting digital inclusiveness. Professor Tang Li emphasized the need to take specific measures in terms of ethical standards, positioning of “human-machine relationships”, application scope, and strengthening technical security guarantees for the application of artificial intelligence in the judiciary, in order to clarify the limitations of AI judicial applications. Professor Xu Duoqi believes that it is necessary to simultaneously eliminate institutional barriers such as cross domain collaboration, division of rights and responsibilities, and standardized application in the process of handling financial cases, while breaking down data barriers, accurately adapting algorithms, and building strong security mechanisms. Professor Zhang Xingmei pointed out that artificial intelligence is currently playing a substantive role as the “fourth party” of intelligence in dispute resolution, but we should also be wary of the alienation of civil justice by technology supremacy. Associate Professor Huang Zemin has established three standards for the intervention of intelligent technology in the judicial field: strong, medium, and weak. Researcher Ji Pingping pointed out that the development of artificial intelligence technology provides a possible path for optimizing the structure of judicial documents. However, due to the dual attributes of factual determination and value judgment inherent in legal reasoning, artificial intelligence is still unable to handle the task of automatically generating documents with standardized creation and reasoning functions.