Epistemic vs. Moral: A Differentiated Approach to Human-Centered AI Ethics

Abstract The issue of human-centeredness represents a crucial challenge in the field of Artificial Intelligence (AI) ethics. To abandon it altogether would mean waiving any attempt at human control over emerging technologies. Clearly, this is an impractical path, given the increasingly prominent role AI plays in decision-making across multiple sectors—from marketing and social life to more sensitive domains such as healthcare and defense. If we also consider the growing sophistication of these tools and their ever-expanding scope, legitimate questions arise regarding whose values, beliefs, and needs are being embedded into these technologies. However, uncritical adherence to human-centeredness risks generating other controversial issues. The development of technologies focused exclusively on human interests may neglect the impact they have on the biosphere and the non-human animals that inhabit the planet. Moreover, depending on how one defines the human, entire groups—such as women, Black people, queer individuals, people with disabilities, and socioeconomically disadvantaged communities—can be marginalized. This dilemma has often been framed as a binary choice: either endorse human-centered approaches and risk anthropocentrism or adopt non-anthropocentric perspectives and risk undermining human agency. This paper challenges this dichotomy by proposing a third strategy: the clear conceptual distinction between human-centeredness and anthropocentrism. Specifically, it argues for an epistemically human-centered approach—grounded in the recognition that our situatedness is inevitably human—that does not preclude moral consideration of non-human beings or ecological systems. Achieving this requires a thorough reexamination of the different conceptual standing of human-centeredness and anthropocentrism, and how their conflation may negatively impact AI ethics. Although achieving this goal is far from unproblematic, clarifying this distinction can help us develop both ethical frameworks and regulatory guidelines. These, in turn, can preserve the centrality of human-centeredness while fostering technologies designed responsibly and inclusively, for the flourishing of all living beings.