The Forgotten Margins of AI Ethics

AI ethics literature often lacks depth on marginalized communities: Despite good intentions, most papers published at major AI ethics conferences (FAccT and AIES) provide only shallow consideration of how AI negatively impacts traditionally marginalized groups, suggesting the field needs to go deeper in understanding real-world harms.

The field is dominated by Western philosophy and mathematical approaches: Current AI ethics work primarily draws from Western canonical philosophy and statistical methods, which may miss important perspectives and fail to adequately address structural power imbalances that affect vulnerable populations.

Real people's experiences are largely absent from the conversation: The paper found that many AI ethics papers don't examine concrete applications, actual use-cases, or people's lived experiences with AI systems—a critical gap for understanding how ethical problems actually manifest in society.

Marginalized groups are often unnamed and invisible in research: Even when papers do address harmed or marginalized communities, they frequently fail to explicitly name which groups are affected, making it harder to track progress and ensure accountability to those communities.

The field needs grounding in real-world context and critical perspectives: To address these gaps, AI ethics would benefit from increased focus on concrete use-cases, lived experiences of affected people, and perspectives from critical data studies that examine how power structures shape AI systems.