Ethical Frameworks for AI
Apply utilitarianism, deontology, and virtue ethics to real AI decisions
Why Philosophical Frameworks Matter Practically
Ethical frameworks aren't just academic exercises. When you face a real decision — should we deploy a model that helps 90% of users but systematically harms 10%? — different frameworks give different answers, and understanding which framework you're implicitly using clarifies the trade-off you're actually making.
Consequentialism (Utilitarianism)
The right action is the one that maximizes overall good. Applied to AI: deploy if total expected benefit outweighs total expected harm. Problem: can justify significant harm to a minority if the majority benefit sufficiently. Amazon's hiring algorithm disadvantaged women but improved overall hiring efficiency — a utilitarian calculus that may feel wrong.
Deontology
Some actions are intrinsically right or wrong regardless of consequences. Applied to AI: certain uses of AI are wrong regardless of their beneficial effects. Automated sentencing without human review violates human dignity even if it's on average more consistent. Privacy violations are wrong even if they enable better health outcomes.
Virtue Ethics
Focus on character: what kind of organization do we want to be? A virtuous organization would be honest about what its AI can and cannot do, humble about uncertainty, and willing to forgo profit when a deployment would cause harm. Useful for shaping organizational culture rather than making specific decisions.