| 摘要: |
| 算法行政能够提升行政效率和决策质量，也带来正当程序虚化、公民权利减损和问责机制模糊化的风险，进而引发了广泛的社会焦虑和不信任。因此，如何构建公众对算法行政的信任，是算法行政面对的重大问题。针对算法行政的法律控制技术，传统上以“禁止性预防标准”为中心，这一风险控制路径过于严格地设定了算法行政的法律保留范围和实体控制边界，在评估方式的科学性与适用范围的场景化两个方面存在缺陷。对此，应从“强风险预防”转向“可容忍风险”的理念，通过理性化的风险沟通与处置机制来动态回应公众预期，避免公众对算法产生不合理的恐慌或盲信。算法行政中的风险不仅源自技术层面，更是社会技术系统运作的结果，涉及政企关系、组织体制、人机互动决策结构与权责分配等问题。因此，算法行政的法律控制不能仅依赖传统的节点式信任构建方式，而应从整体的数字行政生态出发，结合算法行政的社会技术属性，建立系统性的信任机制，实现风险控制与算法赋能的兼容兼顾。 |
| 关键词: 数字行政 风险沟通 风险管理 人工智能 社会技术系统 |
| DOI： |
| 分类号: |
| 基金项目:国家社会科学基金重大项目“社会信用体系的法律保障机制研究”（项目批准号： 21&ZD199）与2025年国家建设高水平大学公派研究生项目（项目编号： 202506010235） |
|
| Legal Approaches to Building Trust in Algorithmic Administration |
|
Huang Zhijie & Wang Xixin
|
|
|
| Abstract: |
| Although algorithmic administration can enhance administrative efficiency and decision-making quality， its implementation also introduces risks such as procedural erosion， impairment of citizens' rights， and obscured accountability mechanisms， leading to widespread societal anxiety. Therefore， how to build public trust in algorithms is a significant issue faced by algorithmic administration. A risk management model centered around a “prohibitive precautionary standard” overly rigidly defines the scope of legal reservation and substantive control boundaries for algorithmic administration， with deficiencies in the scientificity of assessment methods and scenario-specific applicability. Thus， a shift from “strong risk prevention” toward a philosophy of “tolerable risk” is advocated. Rational risk communication and management mechanisms should be employed to dynamically respond to public expectations， preventing unreasonable panic or blind faith in algorithms. Risks in algorithmic administration arise not only from technological factors but also from the operation of socio-technical systems， encompassing government-enterprise relationships， organizational structures， human-machine interactive decision-making frameworks， and the distribution of authority and responsibility. Consequently， legal control of algorithmic administration cannot solely rely on traditional， node-based trust construction methods. Instead， it must start from the holistic digital administrative ecosystem， integrate the socio-technical attributes of algorithmic administration， and build systematic trust mechanisms that balance risk control with algorithm empowerment.
Keywords |
| Key words: Digital Administration, Risk Communication, Risk Management, Artificial Intelligence, Socio-technical System |
