Source
Hooker, SEP “Rule Consequentialism”, LIT-198, read in full in NOTE-160 (the revision as it stood in 2023). SEP entries have no pages, so locations are sections.
What was actually shown
An argued survey. Its formulations are stated exactly and its replies organised by the author who made most of them (Hooker 2000).
- The objection. If rules are chosen for their consequences, the best rule is “maximise, except where…”, with as many exceptions as optimising needs. Then rule-consequentialism prescribes what act-consequentialism prescribes.
- The reply (§§6.2, 8). Codes are ranked by the expected good of “their teaching and acceptance”, not of compliance. Teaching and internalising a code costs more as it gains rules, complexity and demandingness. A code with many exception clauses also invites self-serving misapplication and undermines others’ assurance. So the best code is simpler and less demanding than one optimal act by act, and will sometimes require a sub-optimal act. The toy example (§6.2): Code A has expected value n and internalisation cost −1, Code B has n+5 and cost −7, so A (n−1) beats B (n−2).
- Costs measured from new generations (§6.2). Transition costs are computed for new generations, not from a society’s existing code, to give no weight to unjustified codes and to avoid relativism.
- Incoherence (§8). The view is incoherent only if one assumes an overriding commitment to maximise, which neither the agent’s moral psychology nor the theory contains. The agent’s basic motive can be “to do what is impartially defensible”.
- Rule-worship (§8). A “prevent disaster” rule, triggered only by “a very much larger difference” in expected value, so that it does not re-collapse.
What could have come out otherwise: once teaching costs are counted, the best code could still match act-by-act optimisation everywhere. The toy example shows a case where it does not; the general claim rests on the premises about human nature.
What this does not say
- That the human-nature premises are established. That children start partial to themselves and a few others, that each rule carries a cognitive cost, and the Sidgwickian limit on impartial concern are called “realistic assumptions” without empirical citation (NOTE-160).
- That every formulation escapes collapse. The reply presupposes ranking by teaching and acceptance. It is the teaching-and-acceptance formulation that answers collapse; the prevent-disaster rule answers only rule-worship (NOTE-160, corrections).
- That the incoherence objection “dies”. The headline sentence in §8 outruns the caveat after it: the reply “seems less viable” if rule-consequentialism is argued for from consequentialist first premises.
- Which acceptance formulation is right. The ~90% “overwhelming majority” threshold is admitted to be a compromise, and the teaching formulation is left as a live alternative. The teaching formulation faces the open objection that its teachers cannot be described coherently (§9).
- That the view survives §9. Demandingness in other possible worlds (Mulgan), contingency of justification, the “wrong facts” objection and conflicts among rules are left open.
- Anything about specifying AI behaviour. The mechanism is the cost of teaching children many, complex or demanding rules. An analogy to rule-based versus outcome-based specification of AI is the reader’s own; the entry does not draw it (NOTE-160).
Connections
- Pettigrew’s Wolfian epistemology (LIT-186) transposes Susan Wolf’s rule consequentialism to epistemic norms, and this entry is its background. The entry cites Wolf 2016 in §7, for the route from morality as a social practice with a function, and in §8, on incoherence.
- Sen’s “comprehensive consequentialism”, which counts processes and principles as well as outcomes, is reported in LIT-284 (THEORY-411’s source) and is another non-act consequentialism. Neither work mentions the other.