Autonomous Driving: Regulatory Challenges Raised by Artificial Decision-Making and Tragic Choices
Autonomous Driving: Regulatory Challenges Raised by Artificial Decision-Making and Tragic Choices
复制标题
自动驾驶：人工决策和悲剧性选择带来的监管挑战
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
A. Ungern
中科院分区：
文献类型：
--
作者：
A. Ungern
人工智能意味着人工代理的决策取代了人类的决策。本文研究了这一发展在交通法规的密集监管领域提出的挑战，其中自动驾驶汽车和人类驾驶员同时操作。与其他更开放的法律领域相反，道路交通法通常适合人工智能体的操作，即使国际公约中对人类“驾驶员”的要求仍需要修改。因此，现在应该由立法者和其他监管机构来澄清法律框架。为了确保自动驾驶汽车遵守法律，监管机构必须监督程序员如何将交通规范转化为计算机代码，并且必须为人工事实调查设定可靠性标准。机器学习可能会成为发现事实和有意义的法律遵守的重要方法，如果这个过程受到监督，如果它的结果是可以解释和合理的，那么它与法律的首要地位并不矛盾。此外，监管机构必须解决人为决策造成的偏见，正如本文所阐述的，关于坠机算法的悲剧性生死决策。崩溃算法引发了法律问题，而不仅仅是道德问题，法律理应对此进行监管。在这样做的过程中，监管机构履行了一项来自人权的积极义务，即生命权、平等权和人的尊严，这也应该总体上指导监管选择。更准确地说，碰撞算法本身不侵犯人的尊严，应该反映法律秩序的优先次序，不能利用种族、性别或年龄等个人特征，在事故的潜在受害者之间进行选择。然而，监管机构可以规定死亡人数最小化，具体说明风险领域或诉诸责任作为这些悲剧性决定的相关标准。
Artificial intelligence implies that decision-making by artificial agents replaces human decision-making. This paper examines the challenges this development raises with respect to the densely regulated area of traffic law in which self-driving cars and human drivers operate simultaneously. As opposed to other, more open-textured areas of law, road traffic law is generally suited for the operation of artificial agents even if the requirement of a human “driver” in international conventions still needs to be amended. Thus, it is now up to legislators and other regulatory bodies to clarify the legal framework. In order to ensure legal compliance by self-driving cars, the regulator must supervise how programmers translate the norms of traffic into computer code, and must set standards of reliability for artificial fact-finding. Machine learning, which will likely serve as an important method for fact-finding and meaningful law compliance, is not incompatible with the primacy of law if the process is supervised, and if its results are explainable and justifiable. Furthermore, a regulator must address the biases created by artificial decision-making as the paper illustrates with respect to tragic life-and-death decisions of crash algorithms. Crash algorithms raise legal, not only moral, questions that the law should arguably regulate. In doing so, the regulator fulfills a positive obligation flowing from human rights, i.e. the right to life, equality, and human dignity, which should also generally guide regulatory choices. More precisely, crash algorithms, which do not, as such, violate human dignity, should reflect the priorities of a legal order, and must not use personal characteristics such as race, gender, or age, to chose between potential victims of an accident. The regulator may, however, prescribe death-toll minimization, specify areas of risks or resort to responsibility as a relevant criterion for those tragic decisions.