Contribution

The paper (LIT-509) extends morality-as-cooperation from a list to a generative system. The seven domains of the founding statement (LIT-503) become “elements”. Each pair is hypothesised to combine into a more complex moral concept, and an example of each is located in “the professional and popular literature” (§2). Section 3 counts the possible higher-order and mixed-valence combinations. Section 4 claims that the theory is “the first and only theory to make genuine predictions about what moral phenomena there might be in the world”.

Key insight

Many named virtues and institutions read naturally as two cooperative solutions applied at once:

  • filial piety: deference within the family;
  • honour: heroism expressed through revenge;
  • queuing: division by first possession;
  • arbitration: division settled by deference.

If that holds generally, a small set of cooperative elements could account for a large and culturally variable moral vocabulary. The paper suggests that elements may be innate and universal and molecules acquired and relative (§4).

Assumptions

  • The seven elements are the right basis. The paper’s Table 1 restates them from LIT-503 and the MAC-Q (LIT-496).
  • Conceptual analysis is the method. “Not an exact science”, and “not everyone will be persuaded by all the examples” (§4).
  • The molecules are unordered pairs, so 7!/(2!·5!) = 21 (footnote 5). For each pair the authors hypothesised one candidate, then “tried to find examples … that matched our theoretical expectations, and that had a positive moral valence”.
  • A combination may have unequal parts. Collegiality could be MR² and comradeship M²R (§3), so the number of possible molecules is unbounded.

Key results

  • Table 2, 21 molecules.
    • Kinship with: mutualism, fraternity; exchange, blood revenge; hawk, family pride; dove, filial piety; division, gavelkind; possession, primogeniture.
    • Mutualism with: exchange, friendship; hawk, patriotism; dove, tribute; division, diplomacy; possession, common ownership.
    • Exchange with: hawk, honour; dove, confession; division, turn-taking; possession, restitution.
    • Hawk with: dove, modesty; division, mercy; possession, munificence.
    • Dove with: division, arbitration; possession, mendicance.
    • Division with possession: queuing.
  • Supporting data, where any is cited. Blood revenge is a moral imperative in 38 of 168 societies and legitimate in 90 (Ericksen and Horton 1992). Moral outrage at queue-jumpers is shown experimentally (Milgram et al. 1986). The other cells rest on definitions and quotations.
  • Counts (§3, footnotes 7–8, checked):
    • 2⁷ − 1 = 127 non-empty combinations of the seven, in the proportions 7 + 21 + 35 + 35 + 21 + 7 + 1.
    • 3⁷ − 1 = 2,186 combinations once each element can be absent, positive or negative.
    • The worked ternary example, honour among thieves, 2001100₃ = 1,494, is correct.
  • Revision rule (footnote 9): a new non-cooperative element would force the theory “to be revised more fundamentally”.

Claims

idclaimstrengthsupport
C1Each of the 21 pairwise combinations corresponds to a positively valenced moral conceptweakone example per cell, chosen after the cell was proposed; no protocol, judges or failures reported (“We succeeded in every case”)
C2Morality is a combinatorial systemweakfollows from C1 by analogy with chemistry, DNA, language and music; no test that distinguishes combination from coincidence of description
C3The framework “has the potential to explain all moral ideas, possible and actual”assertionwith unequal proportions allowed, the space is infinite (§3), so fitting it is easy and failing it is hard
C4MAC is “the first and only theory to make genuine predictions about what moral phenomena there might be”weakthe paper also says the theory “doesn’t predict which of these many possibilities will be actualities” (§4)
C5Elements may be innate and universal, molecules acquired and relativeconjectureposed as a question for future work

Concepts

  • Moral element / molecule. One of the seven cooperative domains, and a concept analysed as a combination of two or more of them.
  • Negative element. The vice that matches each element: neglect, betrayal, cheating, cowardice, disrespect, unfairness, theft. Molecules can mix positive and negative elements (“mutiny”).
  • Binary and ternary notation (footnotes 7–8). A seven-digit code for which elements are present, and with what sign.

Connections

  • The founding statement and the MAC-Q (LIT-503, LIT-496). The elements come from the one and the molecule idea from the other’s limitations section (pp. 40–41 of the manuscript).
  • The sixty-society test (LIT-047, NOTE-034). This paper adds no evidence on whether the seven are universally valued, or whether non-cooperative content is moralized. It cites LIT-047 for the claim that each domain “has been considered morally good the world over”, and does not discuss the objections LIT-047’s commentators raised.
  • Purity (LIT-101). Mendicance is read here as humility plus property. It is the one ascetic norm the paper treats. Fitouchi, André and Baumard explain ascetic piety and ritual differently, as practices believed to train self-control. The record’s reading of them (NOTE-132) does not mention poverty vows. Extended to such vows, the two readings would come apart. On this paper’s reading, poverty vows should co-vary with deference and property norms. On LIT-101’s, they should co-vary with other self-control norms such as fasting and chastity. No study in the record compares them.
  • The machine-read replication (LIT-512). It reports filial piety and honour in HRAF paragraphs, as examples rather than counts.

Bearing on the record

There is nothing here for ML practice, and the anthology does not hold the paper. For the morality-as-cooperation thread it shows the programme’s ambition and its method, both at their plainest. It should not be cited as evidence that morality is combinatorial.

Limitations

  • Confirmatory search. No record of where the authors looked, what they rejected, or whether anyone outside the team agreed with the classifications.
  • The elements stretch to fit. Two molecules combine an element with its denial, which the paper’s own scheme (§3) counts as mixed valence:
    • mendicance is “disavowing property”;
    • common ownership is holding property “in common” rather than respecting individual possession. Primogeniture reads birth order as “first-come-first-served” possession, which is an analogy rather than a combination.
  • No discriminating test. The paper names none, and does not say what outcome would show that morality is not combinatorial.
  • Inconsistent naming. The abstract and §3 say “mendicancy”; Table 2 and §2.20 say “mendicance”.

Open questions

  • Given an independent list of moral concepts (a virtue lexicon, the eHRAF Ethics index), how many can blind judges place in the scheme, and with what agreement? How many fit nothing?
  • Is endorsement of a molecule predicted by endorsement of its elements, as §4 asks?
  • Do poverty vows track deference and property norms, or self-control norms?

Corrections

  • none to a seeded skim (there was no seed)
  • The citation. The owner’s citation (Review of Philosophy and Psychology 13:1039–1058, 2022) is correct. The article was accepted 14 March 2021 and published online 11 August 2021, and is in issue 4. An OSF preprint (xnstk) is dated 9 June 2020.