Mark Alfano, Marc Cheong, Oliver Scott Curry (2024), Heliyon 10(6):e25940 — DOI-10.1016/j.heliyon.2024.e25940
Key takeaways
- A dictionary for the seven morality-as-cooperation domains, run over the whole eHRAF Ethics corpus, detects most of the seven in most of 256 societies.
- The tool is valence-blind, and agrees poorly with the hand codes it was validated against, so it extends the presence result of the sixty-society study and leaves its valence result untouched.
Standing in the record
Filed on 2026-10-02 at the owner’s request, with the rest of the morality-as-cooperation programme (ADR-018). It is the programme’s attempt to go beyond the sixty-society hand coding (LIT-047).
NOTE-406 is the close reading of 2026-10-02, and it placed the work Proposed. It bears on NOTE-034’s verdict in three ways, and none of them moves LIT-047 toward Active.
- Valence. The 99.9% positive-valence finding is the result NOTE-034 judged weakly diagnostic, and this paper cannot test it. The authors say the limitation is “probably not a major problem, given that cooperation had a positive moral valence in 99.9% cases”. That leans on the finding in question.
- Counterexamples. The paper’s own spot-check found admired theft among Bedouin, Greek, Pashtun and Wogeo groups. It also found !Kung San who “attach no value to fighting” and a Navajo ethnography reporting no obligation to reciprocate. LIT-047 counted one negative in 962 observations. The authors frame the new cases as “competition between the morals”, which is footnote 3’s move again. These were found by reading for them, and they bear out NOTE-034’s point that the original design could hardly register a negative.
- Regions. All seven domains differ significantly by region (Kruskal–Wallis, p < .001), with small effects (ε² ≤ .055). NOTE-034 judged LIT-047’s “equal frequency across all regions” (its C4) to be absence of evidence; this paper finds a difference.
The comparison NOTE-034 asked for, non-cooperative content coded the same way, is again not run. The authors say it cannot be, because there is “no ground truth coding of Moral Foundations in HRAF”. There is no ML practice in it. It is a word-count method from before large language models, so it is not an anthology candidate.