“We become just by doing just acts, temperate by doing temperate acts, brave by doing brave acts.”
Aristotle, Nicomachean Ethics, Book II, ch. 1 (trans. W. D. Ross)
Virtue Ethics
Aristotle's account of virtue runs through habit, not instruction. You become just the same way you become a builder — by doing the thing repeatedly until it settles into character. But he's careful to add a second condition: a person can perform a just act without being just, the way someone can stumble into a correct sentence without knowing grammar. Real virtue needs the act to come from a firm, settled character, not a one-off correct output.
That distinction is exactly what's at stake in AI trained on human feedback. A model shaped by enough rounds of correction can learn reliably acceptable behavior in situations like the ones it has seen before. What it doesn't have is phronesis: practical judgment for a genuinely novel case nothing in its training resembles. The same test applies to a person leaning on that model. The question was never whether the output looked right. It's what the habit of using it is making of you.
What Virtue Ethics gets wrong
Utilitarians would say this has no answer for scale — knowing your own character doesn't tell you which policy saves more lives, and "what a good person would do" goes quiet exactly when interests genuinely conflict. Stoics would say virtue ethics leans too hard on how you land with other people — being seen as someone of good character — when the more defensible claim is narrower: only your own judgment, not its audience, is actually yours.