Towards a Code of Ethics for Artificial Intelligence
Towards a Code of Ethics for Artificial Intelligence
复制标题
制定人工智能道德准则
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
P. Boddington
中科院分区：
文献类型：
--
作者：
P. Boddington
正如本书标题中包含的“朝向”一词所表明的那样，作者PaulaBoddington并不打算向读者展示人工智能的最终道德准则。相反，她的目标是提供一个路线图，说明如何以及在什么条件下可以形成人工智能的道德准则。作者提供了一个简明的指南，详细介绍了人工智能环境中道德及其编纂的价值和必要性，同时还强调了可能阻碍其制定的特定陷阱和挑战。博丁顿的工作是务实的：它有助于促进最终结果的创造，即道德准则。因此，它应提供方法，使一项困难的程序性工作在实践中更可行。此外，它还应该为非哲学家提供必要的工具，以便在处理人工智能时进行更具价值敏感性的评估。考虑到伦理学和人工智能的深度，需要一个清晰而全面的结构，以使读者能够理解和反思所探索问题的广度。在这方面，作者正确地指出1，起草守则的过程不仅需要了解什么是道德问题，2还需要了解守则和法规的目的。3此外，为了使后者可行且有效，还应该了解谁将参与相关守则的制定、批评和应用。此外，人们应该记住塑造道德和AI构建的背景因素。4通过八章，作者成功地阐明了代码开发的整个生命周期以及可能出现的相关问题。章节遵循自然的进展，即使涵盖了广泛的主题和考虑因素，读者也能够保持监督和理解。特别相关的是评论，“这对人工智能意味着什么”，穿插在整个文本中，有时抽象的道德原则或挑战被应用于说明它们与人工智能的具体相关性。突出了实例或轶事，这增加了主题的可理解性和可读性。人们可能会质疑Boddington将人工智能作为一个整体来解决的雄心。正如作者所指出的，虽然在当今的辩论中无处不在，但人工智能这个术语很复杂，不容易定义。鉴于这种差异，我们如何建立人工智能道德准则的基石？Boddington承认，在辩论中，可能需要关注人工智能的独特特征。6此外，她指出，在原则通常倾向于支持人工智能的情况下，“我们需要考虑的价值问题更加本地化和情境化。7如果人工智能辩论可以从情境化中受益，那么关于创建道德准则的指示也应该遵循情境化的方法？然而，即使从广义上考虑，人工智能也确实提出了特殊的、总体的挑战。例如，Boddington挑战了对“智能”作为人类标志的关注。8不可否认，与人工智能相关的一些挑战与过去、现在和未来其他快速发展的技术一样，例如“炒作”对道德推理的影响。“人工智能中人类代理和推理的扩展，增强和替代是其使用中出现的许多伦理问题的所在地，有时DOI：10.21552/德尔菲/2019/2/11
As indicated by the inclusion of theword ‘towards’ in the title of this book, authorPaulaBoddingtondoes not intend to present the reader with the definitive code of ethics for artificial intelligence. Rather she aims to provide a roadmap on how and under what conditions a code of ethics for AI could take form. Both short and practical in nature, the author offers a concise guide detailing the value of, and need for, ethics and its codification within the AI environment, whilst also highlighting the particular pitfalls and challenges that might impede their formulation. Boddington’swork is pragmatic: it serves towards facilitating the creation of an end-result, ie a code of ethics. Hence, it should provide methods to make a difficult procedural enterprise more workable in practice. It should moreover provide the necessary tools for non-philosophers to make more value-sensitive assessments when dealing with AI. Considering the depth of both ethics and AI, a clear and comprehensive structure is needed in order to enable the reader to understand and reflect upon the breadth of issues explored. In this regard, the author rightfully notes1 that the process of drafting a code not only requires an understanding ofwhat ethical questions are,2 but also what the purpose of codes and regulations might be.3 Furthermore, in order for the latter to be workable and efficient, an understanding should also be created concerning who will be involved in the production, critique and application of the code in question. In addition, one should remain mindful of the contextual factors that shape the construction of both ethics and AI.4 Through eight chapters, the author, with success, sets out to clarify the entire life cycle of code development, and the associated questions that might arise. The chapters follow a natural progression, and even though a wide range of topics and considerations are covered, the reader is able to maintain oversight and comprehension. Particularly relevant is the comment, ‘what thismeans forAI’, interspersed throughout the text in which the sometimes abstract ethical principles or challenges are applied in an effort to illustrate their specific relevance for AI. Practical examples or anecdotes are highlighted, which increases the understandability and readability of the subject matter. One could question Boddington’s ambition to tackle AI as awhole. As the author notes, though omnipresent in current day debates, the term artificial intelligence is complex and not easily defined.5 Moreover, AI covers a broad range of applications, which each raise their own set of particular challenges. Given this disparity, how then do we establish the building blocks towards a code of ethics for artificial intelligence? Boddington acknowledges that within the debate, a focus on distinctive characteristics of AI might be needed.6 In addition, she notes that, where principles often tend to favor AI in general, ‘the value issues we need to consider are much more local and contextualised.’7 If the AI debate could benefit from contextualisation, then perhaps instructions concerning the creation of a code of ethics should also follow a contextualized approach? Yet, even when considered in broad terms, AI does raise particular, overarching challenges. For example, Boddington challenges the focus on ‘intelligence’ as the hallmark of humanity.8 Admittedly, some challenges related to AI are shared with other rapidly evolving technologies, both past, present and future, such as the impact of ‘hype’ on ethical reasoning.9 Furthermore, ethical confrontations in AI might not be all that new.10As the author notes: ‘The extension, enhancement and replacement of human agency and reasoning in AI serve as the loci of many of the ethical issues that arise in its use, sometimes DOI: 10.21552/delphi/2019/2/11