How to build nice robots: Ethics from theory to machine implementation

This thesis investigates morality from a computational perspective by examining how machines can be developed with capacities for moral reasoning and action. It addresses how to overcome interdisciplinary boundaries between moral philosophy and computer science (Paper I), proposes a virtue-theoretic framework for artificial moral cognition (Papers II and III), and highlights issues of using normative ethics in moral machine design (Paper IV). Additionally, it analyzes how ethical decision-making is enabled and constrained by computational resources (Paper V) and explores artificial moral agency – first through an examination of Ishiguro’s Klara and the Sun (Paper VI), and then by proposing a theory that bridges capacity-based and practice-based approaches (Paper VII). The work unfolds along two main threads: Practically, it argues that moral machines should be developed ‘bottom-up’, with careful attention to the moral and non-moral aspects of the human practices in which they are meant to operate. Theoretically, it demonstrates that a computational approach to morality offers exciting opportunities to integrate diverse interdisciplinary insights, thereby enriching our understanding of morality itself. Taken together, this work provides a smorgasbord of challenges and possibilities for moral machines, underscoring the need for interdisciplinary collaboration, technical feasibility, and grounding in human practice.