NOTE-407: Baumard, André & Sperber 2013, mutualistic morality

Contribution

• A mutualistic (+/+) rather than altruistic account of the evolution of morality, based on partner choice instead of partner control (sect. 2.1).
• An argument that partner choice, unlike partner control, pins down the division of gains: fairness as equal marginal benefit per unit invested (sect. 2.1.4).
• A two-step account of a non-instrumental moral sense: calculated fairness first, then intrinsic fairness, selected because it makes a more reliable partner (sect. 2.2.1).
• Mutualistic readings of mutual aid and punishment, and of a body of economic-game results (sects. 2.2–3).

Key insight

The contract that contractualist philosophers describe is not a myth to be explained away. It is the equilibrium of a market for cooperation partners. Where everyone can walk away to someone else, every deal must give each side at least its outside option, and with equal options that means proportionality. A moral sense that tracks this is what partner choice selects.

Assumptions

• Partner choice. Individuals can leave a partner and find another; hunter-gatherer mobility and reputation-based choice are cited as evidence (sect. 2.1.3).
• Equal outside options. The fairness result requires that partners have the same opportunities outside the interaction (p. 63).
• Morality = fairness, by definition. “Here, we use morality in a sense that implies fairness” (note 2). Kin altruism and friendship are cooperation, but “not intrinsically moral” (p. 60).
• Proximate and ultimate kept apart. Selection is for reputation; the psychology is a genuine moral sense (sect. 2.2.1; R3).

Key results

• Formal claim (sect. 2.1.4). With freely chosen partners, biased splits are not evolutionarily stable; payoff should be proportional to invested resources and to talent. The pairwise result is cited from André and Baumard 2011a; for effort and talent “a formal proof will require further modeling” (p. 67).
• Entitlement in dictator games (sect. 3.1). When dictators earned the money, 79% (of $10earners)and70$40 earners) gave nothing, against 19% and 15% without earning (Cherry et al. 2002). When recipients earned it, modal offers matched what they had earned above the $10 floor (Oxoby & Spraggon 2008).
• Merit (sect. 3.2). Modal dictator allocations matched contribution (30 of 73, Frohlich et al. 2004). In Cappelen et al. 2007, about 80% found it fair to reward effort and about 70% to reward talent.
• Mutual aid (sect. 3.3). Framing a game as Maasai osotua (need-based mutual help) lowered transfers and returns (Cronk 2007); Samburu hospitality framing lowered offers to about the share customary for real sharing (Lesorogol 2007).
• Punishment (sect. 3.4). In random-income games, people paid to reduce above-average and raise below-average incomes (Dawes et al. 2007); antisocial punishment in public-goods games is read as retaliation for punishment felt to be unfair (Herrmann et al. 2008).
• Commentary themes.
  • Purity, disgust, harm and the scope of morality. Graham; Ramlakhan & Brook; Rochat & Robbins; Machery & Stich; Atran; Sachdeva, Iliev & Medin.
  • Relationships and need. Clark & Boothby; Warneken; Dunfield & Kuhlmeier.
  • Market and bargaining dynamics. DeScioli; Fessler & Holbrook; Guala; Alvard; Binmore.
  • Altruism and group selection. Gintis; Gill, Packer & Van Bavel; Roberts; Rachlin, Locey & Safin.
  • Mechanism. Ainslie; Ainsworth & Baumeister; Bshary & Raihani; Cova, Deonna & Sander; Tummolini, Scorolli & Borghi; Kirkby, Hinzen & Mikhail; Shaw & Knobe; Cappelen & Tungodden; Bonnefon et al.; Iran-Nejad & Bordbar.
• Response (R4). Two positions. Broad (André, Baumard): kin care, coalitions, hierarchy and purity are moralized only where fairness regulates them; sexual restraint is fairness within monogamy (R4.4, after Weeden et al. 2008). Narrow (Sperber): many such norms rest on other evolved dispositions and are not moral proper.

Claims

id	claim	strength	support
C1	Partner choice with equal outside options selects proportional division	moderate for the pairwise case, weak for effort and talent	cited model (André & Baumard 2011a); “a formal proof will require further modeling” (p. 67)
C2	Partner choice selects a genuine, intrinsic sense of fairness	weak	verbal argument about reliability (sect. 2.2.1)
C3	Economic-game behaviour reflects fairness by entitlement, not altruism	moderate as interpretation	others’ results, reinterpreted; Kirkby et al. note proportionality to “rights” is close to a truism
C4	Group selection is unnecessary, and humans prefer fair to utilitarian outcomes	weak–moderate	R2.4, from studies of non-utilitarian judgment
C5	Kin altruism, friendship and group loyalty are not intrinsically moral	definitional	note 2, p. 60, R4.2
C6	Purity and piety are “orthogonal” to partner value; sexual morality may be fairness	contested between the authors	p. 65 vs R4.4; the authors say they disagree (R4)
C7	Possession is not moral; property rights are moral by mutual advantage	assertion	R3.4, against Tummolini et al.

Concepts

• mutualism — behaviour beneficial to both actor and recipient (+/+), following Hamilton’s classification (p. 60).
• partner control / partner choice — stabilizing cooperation by retaliating against a fixed partner versus by leaving for another (pp. 61–62).
• outside options — what each party could get outside the interaction; they bound acceptable splits (p. 62).
• social selection — selection by the choices of conspecifics, of which partner choice is a case (p. 61).
• fairness — impartiality, or mutual advantage: each party gains equally given what it invested (pp. 63, 77).

Connections

• Morality-as-cooperation (LIT-047, NOTE-034).
  • Shared. Cooperation is the context of morality; game theory is the tool; Trivers and Axelrod are ancestors.
  • Opposed on scope. LIT-047 counts seven cooperative problems as moral domains. This paper counts only fair division, and treats kin and group favouritism as moralized, not moral (note 2; R4.1–R4.2).
  • Opposed on mechanism. LIT-047’s domains come from different games (kin selection, coordination, hawk–dove). This paper’s single mechanism is a market for partners.
  • Data. LIT-047 found its fairness domain in 9 of 60 societies. The narrow coding there (division of disputed resources only) may explain the gap; it is still the least supported of LIT-047’s domains and the whole of this paper’s.
  • Property. Tummolini et al.’s commentary argues that respect for possession is an adaptation independent of fairness, the Maynard Smith line LIT-047 follows. The Response denies possession is moral.
• Moral disciplining (LIT-101, NOTE-132). Fitouchi, André and Baumard carry R4.4’s broad position into a full theory of puritanical norms, with self-control rather than monogamy as the cooperative stake. Graham’s commentary here suggests disgust might cue partner untrustworthiness, close to LIT-101’s later argument.
• MFT. Graham et al. 2013 (LIT-499) and the MFQ-2 paper (LIT-500) call this paper monist. Graham’s commentary asks it to leave economic games and test disgust, individual differences and gossip.
• Relationship regulation (LIT-507). Clark and Boothby’s commentary makes RR’s point: communal and exchange relationships follow different norms. The Response folds need-based help into mutual insurance (R4.1).
• The record’s cooperation thread. Peters and Adamou’s ergodicity solution (LIT-035, NOTE-059) is another mutualistic account, in which sharing pays through variance reduction. It needs no partner market and says nothing about fairness. This paper’s mutual-insurance reading of hunter-gatherer sharing (sect. 2.3.2) appeals to a related benefit of pooling risk, argued from expected utility rather than growth rates.

Bearing on the record

• NOTE-034, C7. This paper shows that a cooperation theory can be narrower than MFT on purpose. “Broader” is not a virtue every cooperative theorist claims; this one defines most of MAC’s domains out of morality proper.
• NOTE-034, C8 (sexual morality is cooperative). R4.4 states the same hypothesis six years before LIT-047’s Reply, with Weeden et al.’s reproductive-strategy data. LIT-101 is its developed form.
• NOTE-034’s open question. The narrow position predicts that purity and in-group norms will be moralized by people but are not “moral proper”, so a blind count of moralized content (the test NOTE-034 proposes) would not settle it for these authors without a definition of “moral”. Machery and Stich’s commentary makes exactly that point.
• No THEORY document cites it. No ML instruction.

Note of 2026-10-16: THEORY-457 now cites LIT-508 as a source.

Limitations

• The evidence is reinterpretation. None of the games was designed to separate mutualistic fairness from altruism or from other accounts; the article says experiments that could falsify it “can and should be devised” (p. 77).
• The core formal result is in another paper; the extension to effort and talent is unproven by the authors’ own statement.
• Defining morality as fairness makes the theory unable to fail on purity, loyalty or hierarchy: what does not fit is not moral (Machery & Stich; Ramlakhan & Brook).
• The commentators’ heterogeneity point stands: in controlled games with shared information, people still split into egalitarians, meritocrats and libertarians (Cappelen & Tungodden), which one shared sense of fairness does not predict without extra assumptions.

Open questions

• Are fairness judgments sensitive to the cost of switching partners, the model’s most direct prediction (DeScioli)?
• Does the broad or the narrow position fit purity better? LIT-101’s vignette mediation results bear on the broad one; a test that separated contamination from puritanical purity would bear on both.
• Why is fairness rare in LIT-047’s ethnographic record if it is the moral sense? A recoding with sharing norms included would answer it.

Corrections

• none to a seeded skim (there was no seed)
• Page range. The target article is pp. 59–78; the full treatment with commentaries and Response runs to p. 122. The citation given for this batch (36(1):59–78) is right for the target article.