This guide shows you how to run a 90-minute structured deliberation on a live AI policy decision, modeled on Aristotle’s account of practical deliberation: reasoning from a shared goal, through the particulars, to a committed decision.
Prerequisites #
- A real, current decision (not a hypothetical) — e.g., a pending procurement, deployment, or rule
- 6–12 participants: decision owner, subject-matter staff, legal/privacy, at least one representative of affected communities, one experienced skeptic
- A one-page case brief circulated 48 hours ahead: the decision, the system, the affected populations, the timeline
The agenda (90 minutes) #
1. Fix the end (10 min) #
Aristotle: deliberation is about means; the end must be in view first. Ask the room to complete one sentence: “This system exists to help citizens ______.” Do not proceed until the sentence is written and agreed. Disagreement here is a finding in itself.
2. Name the extremes (15 min) #
Write the deficiency position and the excess position on the board in the strongest form their advocates would use (see Applying the Mean). Confirm this is a mean-seeking case at all — some uses warrant prohibition, not calibration.
3. Surface the particulars (25 min) #
This is the phronesis core: universals decide nothing without the case in front of you. Work through five prompts: What are the stakes of a wrong decision for one affected person? How reversible are errors, and who would detect them? What does the evidence actually show — in conditions like ours? What happens under the status quo? What would the most experienced caseworker in this domain worry about that isn’t in the briefing?
4. Locate the point and steer from the worse extreme (20 min) #
Propose a calibrated position (scope, safeguards, human decision points). Apply Aristotle’s steering rule (1109a): identify which extreme is more damaging in this case, and set the point conservatively against it.
5. Commit (20 min) #
Deliberation ends in decision, or it was conversation. Record: the chosen position and its reasoning; the named accountable official; the review triggers (what evidence would loosen or tighten the position, and when it will be checked).
Facilitation notes #
- Protect the skeptic. Institutional akrasia thrives where dissent is career-limiting; invite the strongest objection explicitly in stage 3.
- Keep the affected-community voice in the particulars stage — that is where lived knowledge changes outcomes.
- Timebox hard. Aristotle’s deliberator is decisive; open-ended workshops habituate avoidance.
- Repeat on a cadence. One workshop is an event; a monthly practice is habituation (see Virtue Is a Practice).
Output template #
One page: Decision · End served · Extremes named · Particulars considered · Position chosen and why · Accountable official · Review triggers and date. File it where the next deliberation can find it — institutional memory is how the practice compounds.
