Why Philosophy Matters for AI Ethics

The AI ethics discourse has a problem. It is dominated by engineers and policy-makers who are brilliant at building and regulating systems, but who often lack the conceptual tools to ask the right questions. The result is a field that is rich in practical recommendations but poor in philosophical foundations.

Consider the concept of fairness. Computer scientists have proposed dozens of mathematical definitions of algorithmic fairness. It has been shown that many of these definitions are mutually incompatible — that satisfying one necessarily violates another. This is not an engineering failure. It is a philosophical lesson: the concepts we use in ethics are not simple, not formalisable, and not substitutable for each other.

The Missing Foundations

When we ask whether an AI system is fair, we are asking a question about justice. And justice has been debated for three thousand years without resolution. Aristotle distinguished distributive from corrective justice. Rawls proposed the difference principle. Nozick argued for entitlement. Sen emphasised capabilities.

Each of these frameworks leads to different conclusions about what fairness means in practice. An engineer who picks one definition and implements it is making a philosophical choice — whether they know it or not. The danger is that the choice is made implicitly, without examination, and without acknowledging that it is a choice.

The Hard Problem of AI Ethics

The deepest questions in AI ethics are not technical. They are philosophical. What kind of entity could have moral status? What does it mean for a system to suffer? Under what conditions would we owe obligations to an artificial mind?

These are not questions that can be answered by running experiments or measuring outcomes. They require careful conceptual analysis — distinguishing what is possible from what is likely, what is logically coherent from what is merely intuitive, what follows from what.

A Call for Interdisciplinarity

I am not arguing that philosophers should replace engineers in AI ethics. I am arguing that the conversation needs both. Philosophical analysis can identify the hidden assumptions in technical frameworks, clarify the concepts we are using, and expose the value choices embedded in seemingly neutral algorithms.

The stakes are too high for the ethics of AI to be treated as a mere engineering challenge. We need philosophy — not as decoration, but as foundation.