Reconfiguring Diversity and Inclusion for AI Ethics

Diversity and inclusion in corporate AI ethics has shifted away from civil rights foundations. The paper finds that major tech companies (Google, Microsoft, Salesforce) have reframed diversity work using business management language similar to corporate diversity initiatives from the 1980s, rather than grounding it in historical civil rights concerns.

Technical solutions are replacing equity-focused approaches. Companies are focusing on creating "diverse datasets" and using the term "fairness" instead of "equity," which makes AI ethics more actionable for engineers but may oversimplify complex social problems that require deeper contextual understanding.

AI ethics work is becoming disconnected from broader diversity efforts. The paper identifies a critical gap: diversity and inclusion work in AI ethics documentation operates separately from companies' other DEI initiatives and lacks input from subject matter experts who could provide important context for decision-making.

Companies are positioning themselves as "ethics allocators" rather than taking ownership. Rather than taking responsibility for ethical outcomes, these firms are framing their role as distributing ethical considerations to clients and engineers, which may limit accountability for how AI systems actually impact different communities.

Making ethics "tractable" for engineers may sacrifice nuance. While simplifying diversity and inclusion concepts makes them easier for technical teams to implement, this process risks losing important nuances needed to address real-world harms and structural inequalities.