A Capability Approach to AI Ethics

The capability approach shifts focus from AI systems themselves to what they enable people to do. Rather than asking whether an AI tool is "fair" or "transparent" in abstract terms, this framework asks whether the AI preserves or enhances people's meaningful abilities and freedoms—a more practical way to evaluate AI's real-world impact.

This approach clarifies why AI ethics matters in concrete contexts. The paper demonstrates this through medical AI, showing that ethical concerns aren't just theoretical principles but directly relate to whether patients and doctors retain agency, understanding, and control over healthcare decisions.

The capability approach provides actionable guidance for AI designers. Instead of treating ethics as a compliance checkbox, this framework helps developers identify which human capabilities their AI tools might affect—such as a patient's ability to make informed decisions or a doctor's ability to exercise professional judgment—and design accordingly.

Ethics-based auditing of AI systems becomes more systematic and meaningful. The paper argues that auditing medical AI tools through a capability lens allows evaluators to assess whether these systems genuinely support human flourishing in healthcare, rather than just measuring narrow metrics like accuracy or bias reduction.

This approach bridges the gap between ethical theory and practical implementation. By grounding AI ethics in what people can actually do and achieve, the framework makes ethical considerations relevant to engineers, policymakers, and stakeholders who need concrete guidance beyond philosophical debate.