Making Power Explicable in AI: Analyzing, Understanding, and Redirecting Power to Operationalize Ethics in AI Technical Practice

AI ethics fails not because the ideas are wrong, but because power imbalances prevent their implementation. The paper identifies "dysfunction and imbalance of power structures" as the root cause of ineffective AI ethics, rather than flaws in ethical frameworks themselves. This means technical solutions alone cannot fix the problem—the underlying decision-making authority and influence must be addressed.

Current explainable AI (XAI) techniques often serve to legitimize rather than challenge problematic AI systems. Through four empirical cases, the authors demonstrate that making AI systems more "explainable" can actually reinforce unjust power by creating an appearance of transparency without addressing who controls the technology or whose interests it serves.

Affected communities and the public lack countervailing power against those who build and deploy AI. The paper emphasizes that power structures in AI are "dominated by unjust and unchecked power that does not represent the benefits and interests of the public and the most impacted communities," highlighting a fundamental imbalance in who gets to decide how AI is developed and used.

Making power structures visible and accountable is essential to operationalizing ethics. The authors recommend explicitly analyzing and checking power dynamics in AI systems, rather than treating ethics as a purely technical problem. This requires transparency about who holds decision-making authority and how that authority is exercised.

Ethical values must become binding technical standards, not optional add-ons. The paper calls for integrating ethical conduct into the core technical practices, norms, and methods of AI development, suggesting that ethics needs institutional and technical enforcement mechanisms rather than relying on voluntary adoption.