This article explains Aristotle’s claim that virtue is built through habituation — repeated practice — and why that makes one-time ethics statements, signed pledges, and annual training insufficient for Responsible AI in government.
What Aristotle argues (NE Book II) #
Aristotle opens Book II with a decisive claim: moral virtue does not arise in us by nature, nor against nature. We are naturally able to receive the virtues, but we are made good only through habit (1103a). His analogy: people become builders by building and musicians by playing — and in the same way, people become just by doing just actions, moderate by doing moderate actions, courageous by doing courageous actions (1103a-b).
The result of sustained practice is a hexis — a stable, settled disposition of character. And Aristotle adds a crucial test (1105a-b): an action counts as fully virtuous only when the agent (1) acts knowingly, (2) chooses the action for its own sake, and (3) acts from a firm and unchanging character. Doing the right thing once, by accident, or for appearances does not qualify.
He is blunt about the alternative: people who philosophize about virtue without practicing it are like patients who listen carefully to their doctors and then ignore the treatment (1105b).
Why this matters for Responsible AI #
Most organizational AI ethics today is what Aristotle would call knowledge without habituation:
- A published set of AI principles is a statement, not a hexis.
- A one-time ethics review before launch is a single action, not a practice.
- An annual training module produces awareness, which Aristotle explicitly says is not enough.
The Aristotelian prediction — well supported by experience — is that organizations with principles but no practices will behave well under observation and poorly under pressure. Character, individual or institutional, is what you do when the deadline is tomorrow and no one is watching.
The policy translation: build routines, not statements #
If virtue comes from repetition, Responsible AI programs should be designed as recurring practices with cadence, ownership, and consequences:
- Standing ethics case review. A regular (e.g., monthly) session where teams work through a real, current decision — not a hypothetical.
- Pre-deployment challenge as routine. Red-teaming and impact assessment on every significant system, every time, so the muscle stays trained.
- Post-incident learning loops. Blameless reviews after failures, with tracked changes to practice — this is how institutions habituate.
- Practice under low stakes first. Aristotle notes we learn by doing; sandboxes and pilots are the policy equivalent of scales before the concert.
For the full institutional design argument, see Building Virtuous Institutions.
Common objection: “Isn’t this just compliance with extra steps?” #
No — the direction of travel is reversed. Compliance asks “what is the minimum we must document?” Habituation asks “what must we repeatedly do until doing it well is our default?” Compliance produces evidence; practice produces character. A mature program needs both, but only one prevents failures the auditors never imagined.
