According to the chapter, what limitations do AI systems have in achieving moral values, and what is needed to address ethical questions in AI research?

AI systems have intrinsic limitations in achieving moral values: it is difficult to model ethical reasoning about obligations and permissions in machines, ethics involves prescriptions that are not based on physical observations, and many specific ethical questions are difficult even for humans. To address ethical questions in AI research, a well-designed multidisciplinary framework is needed, along with supports such as Ethics as a Service, ethical design training for developers, and repeated ethics education.

The chapter identifies three intrinsic limitations of AI in achieving moral values. First, modeling ethical reasoning about obligations and permissions in machines is challenging. Second, ethics is not solely about social acceptance; it also involves prescriptions that cannot be derived from physical observations. Third, certain ethical questions are difficult for humans themselves, so embedding them in machines appears even more distant. Beyond these, ethical guidelines for AI often fail to deliver their intended effects due to economic incentives and lack of enforcement. The chapter emphasizes that AI researchers, particularly in education, are not equipped to handle ethical questions. Therefore, a well-designed multidisciplinary framework for engaging AI researchers is essential. In addition, the chapter proposes Ethics as a Service, which involves an independent multidisciplinary ethics board, an ethical code, and AI practitioners, and calls for training developers in ethical design, data science ethics, and repeated ethics lessons to build an authentic professional mindset.