Christian List and Philip Pettit (2006), The Southern Journal of Philosophy 44(S1):85–105 (Spindel Conference supplement, March 2006) — DOI-10.1111/j.2041-6962.2006.tb00032.x
Key takeaways
- Group rationality forbids proposition-by-proposition aggregation. If a group is to judge a set of connected propositions completely, consistently and with deductive closure, whatever its rational members think, then its judgment on a proposition cannot in general depend only on the members’ judgments on that proposition. This holds for majority rule (the discursive dilemma, Proposition 1), for any uniform rule (Proposition 2) and for any rule that respects unanimity, even one that varies by proposition (Proposition 3).
- Set-wise supervenience is enough. If the group’s whole set of judgments is a function of the members’ whole sets, rationality is possible: the premise-based procedure and its sequential and distributed variants deliver it (Propositions 4–5). The group may then judge a conclusion in a way that the members’ judgments on that conclusion neither fix nor even enter into.
- Discontinuity without mystery. The group agent is fully individualistic, supervening on what members judge and do under a constitution, and still “may hold judgments that are not directly continuous with its members’ corresponding individual judgments” (abstract). Agency that would arise with no member conceiving of the group, as in Tolstoy’s populations, is set aside: “at this margin of excitement, alas, the intelligibility runs out” (p. 21).
Standing in the record
Filed on 2026-10-04 at the owner’s request, as the lawful substitute for List and Pettit’s Group Agency (LIT-391). The book stays Deferred: OUP’s text is paywalled and neither author has posted it. This paper states, in 22 pages, the supervenience thesis that the book’s Part I (chapter 3, “The Structure of Group Agents”) develops. List’s later paper (LIT-700) cites the two together for proposition-wise supervenience (“List and Pettit 2006 and 2011, ch. 3”). It does not cover the book’s Parts II and III, organisational design and the normative status of group agents. No anthology topic holds it, and it carries no instruction for machine-learning practice.
Carried further by List (LIT-700). “Three kinds of collective attitudes” takes this paper’s proposition-wise and set-wise supervenience as the dividing line between aggregate attitudes, which supervene proposition-wise, and corporate attitudes, which supervene “holistically”, and it cites this paper for the result that divides them (§§2.5, 4.4).
On emergence, against Wimsatt (LIT-707). The owner asked whether group agents and Wimsatt’s aggregativity disagree about emergence. On what is read here, they do not; they use “emergence” for two different things. List and Pettit reject emergence in the strong sense, group agency that “emerge[s] mysteriously, without a clear basis at the level of individuals” (p. 2). Wimsatt, by his own later summary (“Reductionism and its heuristics”, Synthese 2006, §§1–4 and §8 read from the copy his department posts; it is not filed, and the 1997 paper is unread), defines emergence as the failure of aggregativity and holds it compatible with reduction. An aggregative property is invariant under rearrangement, substitution, decomposition and reaggregation of the parts, and scales with their addition and subtraction. In those terms List and Pettit’s results are a proof that a rational group’s judgments are not aggregative:
- their “homogeneity of the supervenience base”, invariance of the group’s judgments under permutation of the members (p. 15), is an invariance- under-rearrangement condition;
- the premise-based procedure keeps it, and the distributed premise-based procedure gives it up (pp. 15–17);
- proposition-wise supervenience, which they show impossible, is close to invariance under decomposing the agenda into single propositions and reaggregating.
So the set-wise, individualistic group agent is emergent in Wimsatt’s sense and not in the sense the authors reject. This comparison is the record’s, not either author’s: neither cites the other, and Wimsatt’s four conditions are known here only from his summary. No relation is declared. A reading of Wimsatt (1997) would check whether his conditions apply to a mapping from judgment profiles to judgments at all.
Neighbours. The paper’s distributed premise-based procedure, subgroups each deciding one premise, is a formal model of distributed cognition of the kind Huebner (LIT-702) uses: his navigation crew integrates routines each “sensitive to some one-dimensional constraint”. List’s 2016 paper (LIT-401) relies on the book’s functionalist definition of agency, of which §2 here is an early statement. The joint-intention accounts §6 cites, Bratman (LIT-488, LIT-471), Tuomela (LIT-476) and Gilbert (LIT-477), are held unread; the paper treats them as describing how members knowingly sustain a group agent, not as accounts of group agency itself.