SGER: Towards Machine Ethics

关于技术和伦理之间的关系的研究主要集中在人类对技术的负责任和不负责任的使用上，少数人对人类应该如何对待机器感兴趣。在所有情况下，只有人类参与了伦理推理。PI们认为，现在是时候为至少一些机器增加道德层面了。认识到涉及机器的行为的伦理后果，以及机器自主性最近和潜在的发展，就有必要这样做。在这个项目中，私人投资经理将通过调查所谓的机器伦理来探索这一层面，机器伦理(与计算机黑客、软件财产问题、隐私问题和其他通常归因于计算机伦理的主题形成对比)涉及机器对人类用户和其他机器的行为的后果。显然，不加约束地依赖机器智能来改变世界可能是危险的。直到最近，机器行为的伦理影响要么是微不足道的(例如，计算器)，要么是只有在人类操作员的监督下才产生的重大影响(例如，机器人汽车组装)。随着我们越来越依赖机器智能，而减少了人类的监督，我们将需要能够指望它们提供一定程度的道德行为。随着机器被赋予更多的责任，如果我们要避免不受欢迎的机器行为，就必须对它们的行为承担同等程度的责任。在这个项目中，私人投资机构将在有限的领域中实施原型系统，一个领域结合了伦理方面的优势，另一个领域学习相互竞争的职责的权重，直到为伦理理论提供决策程序。PI们希望这些调查将加深他们对该领域的理解，以便他们能够撰写一份理由充分的立场文件，阐明机器伦理学的目标并回答反对意见，作为解决这个自20世纪初以来一直令人担忧但几乎没有认真的科学努力的问题的第一步。更广泛的影响：PIS将开发一个机器伦理网站，并将组织一次AAAI机器伦理研讨会，暂定于2005年秋季举行。这项研究的结果将有助于自主智能代理的开发，这些智能代理可以在许多屈从、危险或不受欢迎的任务中取代人类的努力。从长远来看，机器的伦理维度可以用来提醒依赖机器的人，在他们做一些道德上有问题的事情之前，避免可能造成的伤害。这样的能力可能被利用来帮助我们进行道德决策；机器与机器的关系也可能受益，因为新的能力可以为解决资源冲突或预测其他机器的行为提供基础。也许最重要的是，在道德维度的指导下，更完全自主的机器的行为在现实世界中可能比没有道德维度的机器更容易被社会接受。

Research concerning the relationship between technology and ethics has largely focused on responsible and irresponsible use of technology by human beings, with a few people being interested in how human beings ought to treat machines. In all cases, only human beings have engaged in ethical reasoning. The PIs believe the time has come to add an ethical dimension to at least some machines. Recognition of the ethical ramifications of behavior involving machines, and recent and potential developments in machine autonomy, necessitate this. In this project, the PIs will explore this dimension through investigation of what has been called machine ethics, which (in contrast to computer hacking, software property issues, privacy issues and other topics normally ascribed to computer ethics) is concerned with the consequences of behavior of machines towards human users and other machines. Clearly, relying on machine intelligence to effect change in the world without some restraint can be dangerous. Until fairly recently, the ethical impact of a machine's actions has either been negligible (e.g., a calculator) or, when considerable, has only been taken under the supervision of a human operator (e.g., robotic automobile assembly). As we increasingly rely upon machine intelligence with reduced human supervision, we will need to be able to count on a certain level of ethical behavior from them. And as machines are given more responsibility, an equal measure of accountability for their actions must be meted out to them if we are to avoid undesirable machine behavior. In this project, the PIs will implement prototype systems in limited domains, one that incorporates an ethical dimension to its advantage, and another that learns weights of competing duties to the end of providing a decision procedure for an ethical theory. The PIs hope these investigations will sharpen their understanding of the field, so that they can compose a well-argued position paper that clarifies the objectives of and answers objections to machine ethics, as a first step towards dealing with an issue that has been of concern since the beginning of the 20th century but has seen little serious scientific effort. Broader Impacts: The PIs will develop a machine ethics website, and will organize a AAAI Machine Ethics symposium tentatively planned for fall 2005. The outcomes of this research will contribute to the development of autonomous intelligent agents that could be used to replace human effort in many subservient, dangerous, or undesirable tasks. In the longer run, an ethical dimension in machines could be used to alert humans who rely on machines before they do something that is ethically questionable, averting harm that might have been caused otherwise. Such capabilities might be harnessed to assist us in ethical decision-making; machine-machine relationships might also benefit, in that the new capabilities could provide a basis for resolving resource conflicts or predicting behavior of other machines. Perhaps most importantly, the behavior of more fully autonomous machines, guided by an ethical dimension, might be more socially acceptable in real-world environments than that of machines without such a dimension.