This guide gives you a five-step method for using Aristotle’s doctrine of the mean to resolve “innovate vs. regulate” standoffs in real AI policy decisions.
Prerequisites #
- Familiarity with The Golden Mean and Phronesis
- A specific decision in front of you — the method works on cases, not abstractions
Step 0: Confirm this is a mean-seeking case #
Aristotle holds that some acts admit no mean — they are wrong as such (1107a). Before calibrating, ask: does this use of AI treat citizens as objects to be manipulated, strip them of legal agency, or punish them for who they are rather than what they did? If yes, the answer is prohibition, not calibration. Only proceed if the use is legitimate in kind and the question is one of degree.
The five steps #
1. Name the decision precisely #
Not “should we allow facial recognition?” but “should the transit authority deploy live facial recognition for fare enforcement at all stations?” The mean is relative to particulars; vague questions produce vague means.
2. Name the excess and the deficiency actually present in the debate #
Write both vices down in the words their advocates use. Example — generative AI in benefits casework: deficiency = banning caseworkers from any AI drafting assistance, guaranteeing continued backlogs that themselves harm claimants; excess = letting an unreviewed model draft eligibility determinations. Naming both extremes as vices reframes the debate: the question is no longer “for or against” but “where between.”
3. List the particulars that locate the mean “relative to us” #
Aristotle’s phrase means the right point depends on circumstances (1106a-b). The particulars that matter most in AI cases:
- Stakes — what does a wrong decision cost the affected person?
- Reversibility — can errors be caught and corrected quickly?
- Affected populations — are those bearing the risk able to detect and contest errors?
- Evidence base — has this system been tested in conditions like ours?
- Alternatives — what happens under the status quo? (Inaction has victims too.)
4. Locate the mean deliberatively #
The mean is where the phronimos — the person of practical wisdom — would set it (1107a). Institutionally, that means a deliberative body with relevant experience, including people from affected communities, working through steps 1–3 together. Aristotle also advises steering away from the worse extreme (1109a): if the harms of the excess are severe and irreversible while the harms of the deficiency are delays, set the point conservatively — and vice versa.
5. Set review triggers before you commit #
Because the mean is relative to circumstances, it moves when circumstances move. Write down in advance what evidence would justify loosening or tightening: error rates by population, incident counts, independent audit results. This converts a one-time compromise into a governed calibration.
Verify it worked #
A successful application produces a decision memo containing: the named extremes, the particulars considered, the chosen point with reasoning, the accountable official, and the review triggers. If any element is missing, the exercise collapsed back into bargaining.
