Henry Dobson

Henry Dobson (PhD in Philosophy, 2024), A Common Morality Approach for AI Ethics

Over the past ten years, more than 100 AI ethics documents have been published, each with their own unique framework of ethics principles. Recent research has shown, however, that these principles have been largely ineffective when it comes to providing moral guidance within the design and development of AI technology. What all these documents reveal philosophically is that the prevailing approach in AI ethics is the approach known as principlism. Principlism also happens to be the primary approach in bioethics. I start this thesis, therefore, by tracing the origins and history of principlism, which is found within the history of bioethics.

In Chapter 2, I look at four different philosophical problems with principlism as an approach in applied ethics. In Chapter 3, I look specifically at one AI ethics document which has been developed explicitly around the four bioethics principles and, therefore, in accordance with principlism. I argue that the same four philosophical problems with principlism also pose problems for AI ethics as well. To overcome the problems in bioethics, some bioethicists have adopted the theory of common morality for grounding particular moral rules and principles and for developing a more systematic approach in applied and professional ethics. After examining one prominent version of the common morality in Chapter 4, I take the same approach by adopting the theory of common morality for AI ethics.

In Chapter 5, I develop a theoretical framework that is designed specifically for the purposes of AI ethics. I then apply this framework to two different cases in AI ethics. Thus, Chapters 5 and 6 will be where I present a common morality approach to AI ethics. I conclude this thesis in Chapter 7 by critiquing the common morality approach against the four philosophical problems raised in Chapter 2. Ultimately, I argue that AI ethics ought to move beyond the approach of principlism and adopt the theory of common morality instead. By doing so, I believe AI ethics can overcome not only some of the philosophical problems facing principlism but can also be further developed into a more philosophically robust and systematic moral theory that can be used for addressing the ethics of artificial intelligence.

Supervisors: Dr Andrew Alexandra and Dr Dana Goswick