This article scales Aristotle’s account of character formation from individuals to institutions: how public agencies develop — or fail to develop — the settled dispositions that make Responsible AI reliable rather than performative.
The Aristotelian premise: law shapes character #
At the close of the Ethics, Aristotle concedes that arguments alone rarely make people good; most people are shaped by habit and by law, which is why legislation and character formation are inseparable (Book X, chapter 9, 1179b–1180a). Institutions are habituation machines: their incentives, routines, and role models determine what their people repeatedly do — and therefore what they become.
The mechanisms of institutional habituation #
- Leadership modeling. Staff learn what the institution actually values by watching what leaders reward, tolerate, and do under pressure. A director who overrides a risk review to hit a launch date has taught a course no ethics training can offset.
- Incentive alignment. If promotion tracks shipping speed and never tracks judgment quality, the agency is habituating recklessness by design. Add responsible-deployment outcomes to performance criteria for the officials who own systems.
- Practice routines. Recurring ethics case reviews, red-team drills, and incident retrospectives are the institutional equivalent of Aristotle’s repeated actions (see Virtue Is a Practice).
- Institutional memory. A hexis is a stable disposition. Documented precedents, decision memos, and searchable incident histories let character survive staff turnover.
Two institutional vices to watch for #
Ethics-washing — the semblance of virtue. Aristotle insists that genuinely virtuous action flows from a firm character, not from doing the right-looking thing for appearances (1105a-b). An agency with a beautiful AI principles page and no practice cadence is performing virtue, not possessing it. The test is behavioral: what happened the last time a deadline collided with a review?
Institutional akrasia — knowing better, doing worse. Book VII analyzes akrasia: acting against one’s own better judgment under the pull of appetite or pressure. Its institutional form is familiar — the agency that knew the system underperformed for a subgroup and deployed anyway because the political timeline demanded it. Countermeasures are structural, not exhortative: hard gates that cannot be waived informally, and escalation rights for staff who raise concerns.
Trust as infrastructure #
Books VIII–IX treat friendship — including civic friendship, the goodwill that holds a political community together — as something lawgivers care about even more than justice, because communities run on trust. Public trust in government AI is not a communications problem to be managed after deployment; it is an asset built the same way character is: by repeated trustworthy action, visibly performed. Every honored appeal right and every published audit is a deposit.
A maturity ladder #
Compliance (we document what the rules require) → Capability (we can detect and correct our own failures) → Character (doing this well is who we are, and it holds under pressure). Most agencies are at stage one. The mechanisms above are the climb.
