演算法解釋權與演算法治理路徑研究

面對日益緊迫的演算法治理需求，演算法解釋權被提出，對用戶和相關個體的自治性加以尊重，為用戶和相關個體的技術性正當程式權利奠定行使基礎、避免演算法危害成本外化和彌散。歐盟《通用資料保護條例》在立法層面構建了限制和弱化版本的演算法解釋權，但通過資料主體權利和資料保護影響評估制度在法律實施層面對其加以補強。但其仍然存在制度構造不足、語句模糊不清、適用範圍有限等問題。在構建本土化的演算法解釋權時，應當明晰演算法解釋權在演算法治理中的地位和功用，厘清其行使要件和核心內容，以社會嵌入性和應用領域為基準探索精准化、場景化的衡量機制，打造內外兼具的技術協同治理機制。

With the increasingly urgent need for algorithm governance, the right to explanation of automated decision-making has been put forward, which has become a ground for empowering users and relevant individuals with respect to their autonomy, the exercise of their technological due process rights, and the avoidance of externalization and shifting of costs and harm caused by algorithm operations. The General Data Protection Regulation(GDPR)designed the right to explanation of automated decision making in a limited and weakened manner, but it created a combined and reinforced version for algorithm explanations by recognizing various rights of both the data subject and the data protection impact assessment mechanism. Nevertheless, the GDPR scheme has several limitations, such as the relevant provisions being made in an incomplete structure, leaving ambiguity and inappropriate limiting of the application scope. When establishing a localized version of the right to an explanation of algorithms, the position and function for algorithm governance should be properly clarified. The major elements and contents of the right should be fully identified, and in accordance with the degree of social embeddedness and specific application fields. An accurate and scenario-based scheme for explanatory measurement should also be introduced, followed by an integrated approach to governance built on collaboration between internal and external actors.