Empirical AI Ethics: Reconfiguring Ethics towards a Situated, Plural, and Transformative Approach

Top-down AI ethics frameworks often fail communities most affected by AI systems. The paper argues that mainstream AI ethics relies on abstract, principle-driven approaches that don't account for the "situated realities of diverse communities," meaning they miss how AI actually impacts real people in specific contexts.

"Ethics washing" is a real problem in corporate AI ethics. The paper identifies that AI ethics frequently serves as a public relations tool for companies rather than preventing actual harm, with critics arguing the field has become "complicit in sustaining harmful systems rather than challenging or transforming them."

Current AI ethics approaches miss the bigger picture of power and practice. Both top-down principle-based models and bottom-up implementation-focused approaches "fall short in addressing the complexities of AI as a socio-technical assemblage, embedded in practice and entangled with power"—meaning they don't adequately consider how AI systems are intertwined with existing social inequalities.

AI ethics needs to be grounded in real-world evidence, not just abstract principles. The paper calls for "a shift in foundations: from top-down abstraction to empirical grounding," suggesting that ethical frameworks should be built from studying actual impacts rather than theoretical ideals.

Western perspectives shouldn't dominate AI ethics globally. The paper advocates for "pluralisation: moving beyond Western-centric frameworks toward a multiplicity" of approaches, recognizing that different communities may have different values and needs regarding AI governance.