人工智能的可解释性与AI的法律责任问题研究

人工智能(AI)作为类人类智能,无论我们是否赋予其主体资格,在解决其法律责任问题时,都必须对其行为进行解释,为此,探讨人工智能的法律责任问题,应该基于人工智能行为的可解释性的全新路径来推进,而不是纠缠于当下学界关于人工智能主体地位与法律责任的各种主体论与责任理论。人工智能的可解释性,亦即解释人工智能如何在大数据的基础上进行算法决策。然而,在AI领域,虽然以深度学习为代表的人工智能技术已取得了令人瞩目的成就,但如何确保以非技术性的方式向最终用户和其他利益相关方解释算法决策以及任何驱动这些决策的数据,仍是一个无法得到解决的难题,人工智能"黑箱"释明难题决定了人工智能行为的不可解释性。法律责任的本质是答责,不具有可解释性的人工智能不能自我答责,因此其无法承担法律责任;法律责任的目的是预防,不具有可解释性的人工智能无法实现法律责任的预防目的。人工智能法学研究的下一个前沿问题,是人工智能的可解释性问题。

On the Explainability and Legal Liability of Artificial Intelligence

Artificial intelligence(AI)is a kind of intelligence that is similar to human,and in solving the problem of its legal liability,we must explain its behavior,regardless of whether or not to give it subject qualification.Therefore,the discussion on the legal liability of AI should be based on the new path of the explainable AI behavior(XAI),instead of being entangled in various subject theories and liability theories about the subject status and legal liability of AI in the current academic circles.The explainability of AI is to explain how AI makes algorithmic decisions on the basis of big data.However,in the field of AI,although AI technology represented by deep learning has made remarkable achievements,how to ensure that algorithmic decisions and any data driving these decisions are explained to end users and other stakeholders in a non-technical way is an unsolvable problem.The interpretation problem of intelligent“black box”determines the unexplainability of AI behavior.The essence of legal liability is accountability.AI without explainability cannot answer for itself and therefore cannot bear legal liability.The purpose of legal liability is to prevent,and unexplainable AI cannot achieve the purpose of preventing legal liability.The next frontier of legal research on AI is the explainability of AI.