人工智慧風潮下公共服務制度設計策略與法規調適作為之研究

Great Taste and Less Filling? A Review of Institutional Design Strategies and Regulatory Adjustment Action for Public Service through Artificial Intelligent Application

10.29865/PAP.202406_(78).0004

張濱璿(Brian Pin-Hsuan Chang)；陳敦源(Don-Yun Chen)；廖洲棚(Calvin Zhou-Peng Liao)；黃心怡(Hsini Huang)；王千文(Chian-Wen Wang)；張瀠心(Ying-Hsin Chang)；陳郁函(Yu-Han Chen)

制度設計 ； 法規調適 ； 人工智慧 ； 公共服務 ； 演算法 ； institutional design ； regulatory adjustment ； artificial intelligent ； public service ； algorithm

行政暨政策學報

78期（2024 / 06 / 01）

95 - 126

繁體中文;英文

背景：政府運用人工智慧（AI）技術的數位轉型過程中，法規制度建立及風險控管措施，都需前瞻的政策規劃，值得全面探究思考。方法：本文參採制度分析與發展框架（Institutional Analysis and Development framework, IAD）結合管制影響評估（Regulatory Impact Assessment, RIA）理論的內涵，藉由法學比較分析法彙整上開理論之所需要件，蒐集歐盟、OECD等國際組織及美、英等先進國家，對應用AI於公共服務時，所應考量之問題、遭遇困難、制度解決方法與策略等進行彙整，尤其對人權保障所考量議題，以及政策執行機制操作方式深入整理，並進行國內外制度落差比較分析，討論可能的法規調適作為。結果：國外於AI監理層級單位的專責、領域的多元差異、共通執行流程，都有一定的成熟發展路徑，由初期的政策白皮書，發展至較為明確的指引，最後成為具體法規草案，可為我國借鏡。經由比較，本文以四個原則面向分析：倫理導向、風險回饋、組織監管及循證決策，提出我國未來公共服務導入AI之法規調適作為建議。研究發現及建議：從倫理導向切入，各國共通性原則皆立基於「以人為本」之人權保障；在風險層面，需以不同服務風險導向進行不同管制措施；組織層面，建議需分別有負責整合協調及監督治理之組織；決策方面，建議以倫理價值為基礎之具體規範指引，使AI應用之發展保有彈性。

Background: During the government digital transformation process, using artificial intelligence (AI) technology, the establishment of regulations and risk control measures all require forward-looking policy planning and worth to be fully explored and considered. Methods: Starting from the theory of Institutional Analysis and Development framework (IAD) combined with Regulatory Impact Assessment (RIA), this study uses the approach of literature comparative analysis to review the experiences of international organizations and advanced countries. We will propose problems that should be considered, difficulties encountered, regulatory system solutions and strategies, etc. When it comes to applying AI in public services, we especially focusing on human rights protection issues and the policy implementation mechanism. Result: There are certain mature development paths for the AI supervision mechanisms, field diversity, and the executive processes in most of the developed countries. From the beginning policy white papers to relatively clear guidelines and finally a specific draft of regulations, the experiences of other countries can serve as our examples. In the comparative study, we applying the following four principles: ethical orientation, risk response, organization supervision and evidence-based policies. Finally, we propose methods to introduce AI in future public services. Research findings and suggestions: From the perspective of ethics, the common principles of all countries are based on the idea of "people-centered" human rights protection. At the risk level, different control measures should be implemented according to different service risk orientations. From the organizational level, it is suggested that there should be separate organizations responsible for integration, coordination and supervision. In terms of decision-making, it is recommended to provide specific normative guidelines based on ethical values, so that the development of AI applications remains flexible.