Towards machine ethics: Human and computational moral judgements of ethical decisions made by AI (Ref: 4659)

随着《黑客帝国》和《银翼杀手》等电影的普及，难怪这么多人担心人工智能（AI）接管（Cellan-Jones，2014）。随着人工智能越来越多地融入医疗保健和刑事司法系统，这种恐惧正在不断增加，人工智能的前景是能够做出伦理决策，优先考虑哪些患者进行分类（Terwiesch等人，2021），或决定康复或惩罚（Dressel & Farid.，2018年）。尽管如此，人工智能对社会发展至关重要，例如，通过促进早期干预改善医疗保健，从而降低医疗保健成本（Syed-Abdul，2018）和减少刑事司法系统的偏见（Satell，2018）。显然，了解人工智能是否可以被教导做出符合我们道德原则的道德决定，以及我们是否可以学会信任这些决定，既关键又紧迫（Bostrom，2014）。因此，这项博士研究旨在通过理解人们对人工智能道德决策的道德判断和检查可以产生这种判断的计算模型来填补这一空白。

With films like the Matrix and Blade Runner popularising cinema, it is no wonder so many fear artificial intelligence (AI) taking over (Cellan-Jones, 2014). This fear is ever-increasing the more AI becomes integrated into healthcare and criminal justice systems, with prospects of AI being able to make ethical decisions for which patients to prioritise in triage (Terwiesch et al., 2021), or deciding between rehabilitation or punishment (Dressel & Farid., 2018). Despite this, AI is critical to societal development, for example, by improving healthcare by promoting early intervention and, in turn, reducing healthcare costs (Syed-Abdul, 2018) and reducing bias in the criminal justice system (Satell, 2018). Evidently, understanding whether AI can be taught to make ethical decisions aligned with our moral principles, and if we can learn to trust those decisions, is both critical and urgent (Bostrom, 2014). Thus, this doctoral research aims to fill this gap by examining through both understanding people's moral judgments of AI's ethical decision-making and examining computational models that can generate such judgments.