Crunching the numbers. Responsibly.
Intent is a key factor for moral judgments of harm, but less of a factor for purity violations.
convergent domain-by-intent interactions across selected vignettes, with narrow populations and domain operationalizations limiting scope
Replication outlook: uncertain
Based on the agent’s innocent intent, participants judged accidental harms less morally wrong than accidental incest.
direct comparisons and consistent interaction patterns across Experiments 1A, 1B, 2, and 3
Replication outlook: favorable
Based on the agent’s guilty intent, participants judged failed attempts to harm more morally wrong than failed attempts to commit incest.
direct Experiment 4 comparison with small cells and a localized effect-size reporting error
Replication outlook: uncertain
These patterns were specific to moral judgments versus judgments of the agent’s control, knowledge, or intent, the action’s overall emotional salience, or participants’ ratings of disgust.
dedicated alternative-rating experiments support the distinction, though single-item measurement limits precision
Derived from the full evaluation — not a separate score.
Strengths
Experiments 1–5 repeatedly test the intent-by-domain pattern through accidents, failed attempts, perspective changes, and different purity examples. This programmatic structure supports a bounded but meaningful theoretical contribution.
↳ Experiments 1–5; General Discussion §8
Experiments 2 and 3 measure emotional salience, control, knowledge, perceived intent, and disgust rather than merely asserting that these factors cannot explain the moral-judgment pattern.
↳ Results §§4.2 and 5.2; Table 1
The Discussion labels the proposed functional accounts as extending beyond the behavioral data and identifies perspective, cultural variation, and additional moral domains as questions for later work.
↳ General Discussion §8, pp.210–212
Limitations
All experiments use self-selected MTurk participants and a small set of harm, incest, and taboo-ingestion vignettes. This limits conclusions about other populations, cultures, moral domains, and less salient scenarios.
↳ Methods §§2.1, 4.1, 5.1, 6.1, and 7.1; General Discussion §8
Moral wrongness and alternative constructs are measured with individual rating questions, without reliability estimates or validated multi-item scales. Table 1 also contains boundary-valued distributions, including no variation for accidental-harm knowledge ratings.
↳ Methods §§2.1, 4.1, and 5.1; Table 1
Experiment 4 pairs F(1,76)=65.8 with an incompatible partial eta-squared of 0.16, while Experiment 5 does not clearly explain the observations behind df(1,111) after reporting 57 analyzed new participants.
↳ Results §6.2, p.209; Methods §7.1 and Results §7.2, pp.209–210
The score is driven by a coherent five-experiment program in which randomized vignette assignments repeatedly recover the predicted domain-by-intent interaction. Experiments 2 and 3 add methodological value by directly examining emotional salience, perceived control, knowledge, intent, and disgust. The assessment remains below the top band because all studies use narrow Western online samples and single-item judgments, while later comparisons include small cells and two localized reporting ambiguities. The Discussion is notably well calibrated, explicitly separating the behavioral findings from broader functional accounts and identifying perspective and cultural replication as future work.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Strong3.9
The coordinated experiments establish a meaningful comparative account of how intent is weighted in harm and selected purity judgments. The advance remains bounded to two domains, a small scenario set, and limited later-study comparisons.
Intentions are assigned more weight for moral judgments of harm violations, like assault, compared to purity violations, like incest.
Randomized factorial vignette assignments, repeated replications, counterbalanced question order, and direct tests of alternative explanations support the central inference. Convenience sampling, single-item ratings, small later comparisons, and localized reporting inconsistencies reduce methodological breadth and transparency.
Each experiment is introduced as a test of a defined prediction or alternative explanation, producing a readily traceable cumulative argument. The Discussion clearly distinguishes behavioral results from broader functional speculation.
Of course, these theoretical accounts go far beyond the current behavioral data.
The paper substantively connects the findings to work on intent, moral domains, disgust, perspective, and cultural variation. Cross-cultural and scenario-scope boundaries are acknowledged, although their implications for the present Western MTurk evidence are not developed in depth.
Lower confidence on Methodological Rigour — domain match limited.
Caveats4 of 4 checks
Two localized reporting inconsistencies affect effect-size and sample-accounting transparency but do not overturn the replicated principal pattern. The substantive test statistics remain plausible and the concerns are confined mainly to Experiments 4 and 5.
The paper reports multiple online human-participant experiments without an explicit ethics-approval or consent statement. No vulnerable population, coercion, conflict, registration, or questionable availability claim was identified.
Flags: 0 declared / 5 total
63 of 64 checkable references verified
65 references in manuscript 1 have no canonical index record — counted, but not index-checkable 1 reference confirmed by manual review
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Low2.5
Relevance does not apply because the article frames a basic question about moral cognition and identifies no practitioner, organization, or operational decision process as a user.
The findings characterize a judgment pattern under controlled vignette conditions but are not validated in legal, policy, educational, or other operational settings.
The interaction recurs across experiments, multiple scenarios, and second- versus third-person framing. Transfer beyond Western MTurk participants and the selected harm, incest, and ingestion examples remains uncertain.
The work connects to an active moral-domain literature and identifies concrete extensions involving perspective, culture, fairness, loyalty, authority, and dishonesty.
Lower confidence on Readiness, Generalizability — domain match limited.
AI-generated, human-governed. Something look off? Contact us to request a review.