演算法自動化決策之請求解釋權初探：從歐盟GDPR到AI Act

Exploring the“Right to Explanation”of Automated Decision-Making: From the EU GDPR to the AI Act

人工智慧的不透明與難以問責之特性，一直是規範論辯的核心，而提供演算法解釋，則被視為提升透明度與強化問責之重要途徑。自2018年歐盟《一般資料保護規則》（General Data Protection Regulation，GDPR）施行以來，諸多文獻圍繞於GDPR所謂的「請求解釋權」（right to an explanation）分析，認為該項權利或可為演算法的差別待遇提供救濟途徑。然而，關於GDPR是否真正賦予資料主體此一權利以及該權利應如何行使，學界見解分歧。
本文以GDPR以及歐盟《人工智慧法》（The Artificial Intelligence Act，AI Act）為基礎，輔以近年荷蘭阿姆斯特丹地區法院、上訴法院及歐盟法院（CJEU）之相關判決，重新檢視「請求解釋權」於歐盟法體系下之定位與實踐可能性。本文進一步探討如何建構具透明性與有效問責之演算法自動化決策體系，以落實對於資料主體之實質保障。

Providing algorithm explanations is an essential measure to solve the opacity and unaccountability problem of artificial intelligence (AI). Since the General Data Protection Regulation (GDPR) came into force in the European Union (EU) in 2018, many discussions about how the GDPR requires algorithms to have sufficient transparency and accountability have revolved around the debates on the ''right to explanation.'' By asking the data controller to reveal how the algorithms come to a decision, the ''right to explanation'' is believed to help address the automatic decision-making bias and support the remedies of such bias. However, due to the ambiguity of the wording in the GDPR, the theory of whether there is a ''right to explanation'' under the GDPR and how this right should be exercised has not yet been conclusive.
Recently, developments in both practical and doctrinal aspects provide a new basis for discussing the norms and practices of the right to explanation. In 2021, the Amsterdam District Court made a judgment that was seen as the very first recognition of the right to explanation by the courts in the EU. Subsequently, the Court of Justice of the European Union (CJEU)’s rulings in the SCHUFA case and the Dun & Bradstreet Austria case further deepened the theoretical development regarding the right to request an explanation. Furthermore, the Artificial Intelligence Act (AI Act), passed by the European Union in 2024, clearly includes the right to explanation.
The development above appears to be an opportunity to examine the efficacy of the right to explanation and further expand the discussion of the safeguard mechanisms of automatic decision-making. In this context, this article will re-examine the right to explanation in the EU laws, and analyse how to protect the rights of data subjects by ensuring transparency in an algorithmic automated decision-making system. This article examines the dilemma of the GDPR’s right to explanation by analyzing case law, including judgments from the Amsterdam court and the CJEU, engaging with the European Data Protection Board’s discussions, and comparing relevant provisions in the GDPR and the AI Act, with the aim of clarifying its implementation and strengthening data subject protection in the context of automated decision-making.
This article concludes that the right to explanation has practical and doctrinal difficulties. Therefore, to achieve appropriate protection for data subjects, the establishment of personal rights and systematic supervision are necessary.