This paper investigates the moral implications of algorithmic decision-making in machine ethics. It challenges the assumption that even well-functioning ethical intelligent agents (AI) necessarily meet the requirements of autonomous moral agents like humans. Three first-order normative theories—Kantianism, act utilitarianism, and virtue ethics—are examined to illustrate the difficulties in creating both implicit and explicit ethical agents. By comparing these theories and highlighting the differences between calculation and moral estimation, the paper questions the feasibility of both 'strong' (human-level moral cognition) and 'weak' (pre-programmed ethical behavior) moral AI scenarios.

Silviya Serafimova