人工智能的道德性何以實現？──基於原則導向治理的法治進路

How is the Morality of Artificial Intelligence Possible? --An Legal Approach Based on Principle-oriented Governance

人工智能是模擬、延伸和拓展人類智能的新興技術，其道德性的實現尤為重要。人工智能的道德失範既是一種客觀的現象描述，也是一種主觀的價值判斷，既有積極效應，也有消極效應。從全球人工智能道德性實現的實然應對來看，人工智能道德準則的探索呈現出技術主義、降維虛化、內核分歧的特點，並不能為道德性的實現提供可行方案。這一困境的根本原因在於人工智能技術性與道德性之間的衝突難以調和，而以“技術－法治－人文”為范式進路，探索技術性與道德性的激勵相容之道，是這一問題的關鍵所在。對此，引入原則導向治理理論，可以在人工智能道德性實現的理想狀態與現實規範之間創造出包容審慎的制度空間，構建由核心原則框架與工具選擇共同組成的人工智能原則導向治理體系，實現人工智能道德性在道德價值、技術要求和法律語言之間的“轉錄”。

Artificial Intelligence is a newly emerging technology which stimulates, extends and expands human intelligence, thus its moral practice is particularly important. The moral anomie of artificial intelligence is not only an objective description of phenomena, but also a subjective judgment of value, with both positive and negative effects. From the practical response of the moral practice of global artificial intelligence, the exploration on the moral principles of Artificial Intelligence shows technicism, and the phenomenon of suspension and blur appears in the process of transformation from high-level principles to specific norms. Also some of the principles are inconsistent with each other. The essence of the moral anomie of artificial intelligence is the interest imbalance brought by the game relationship between innovation and safety in the technological development. While the approach of ''technology-rule of law-humanity'' to explore the incentive compatibility between technology and morality should be the necessary and key to the moral practice of artificial intelligence. Based on this, the introduction of the principle-oriented governance theory creates an inclusive and prudent institutional space between ideal orientation of moral practice and realistic norms, and set up a system of principle-oriented governance consists of structure of core principle and the choice of tools to transcribe the morality of Artificial Intelligence among moral values, technological requirements and legal language.