**P1 (Empirically limited moral input).** Across languages and cultures, children encounter only a sparse, fragmentary stream of moral examples—speech, observation, and feedback that is *orders of magnitude* smaller than the space of possible adult moral judgments.
**P2 (Early & cross-cultural moral convergence).** Nevertheless, infants and young children everywhere display the same robust expectations about harm, fairness, intent, reciprocity, and helping vs. hindering; later cultural learning elaborates these cores but never abolishes them.
**P3 (Functional and practical necessity).** Stable cooperation and coherent everyday reasoning *require* norms that track harm, fairness, and intent; even self-declared relativists act on these shared intuitions in real life.
**P4 (Poverty-of-stimulus / learnability constraint).** Formal learning theory shows that when learners infer rules that *outstrip* the data available, they must contribute structured priors of their own; without such priors, the universal patterns in P2 would be unlearnable from the impoverished input in P1.
**P5 (Computational modeling evidence).** State-of-the-art domain-general learners (large language models trained on corpora vastly larger than any child’s input) fail systematic moral-reasoning tests unless endowed with explicit moral priors or alignment rules—confirming that data volume alone is insufficient.
**Conclusion (C).** Humans therefore inherit an innate, species-wide moral-cognitive substrate—a constrained *universal moral grammar*. Cultural differences unfold only within its fixed boundaries, so unconstrained moral relativism is false.
**Example With Argument**
1. **P1 — Moral Core** Across cultures and in early infancy, humans share a rule forbidding *unprovoked lethal harm* (part of the universal moral grammar established in the unified argument).
2. **P2 — Functional Necessity** Social cooperation and coherent practical reasoning are impossible if members can be arbitrarily killed; therefore any society that permits unprovoked killing collapses essential trust and coordination.
3. **P3 — Learnability Fact** Children reliably acquire the “no unjustified killing” rule despite meager moral input, which (by the poverty-of-stimulus principle) implies the rule is written into our innate priors.
4. **P4 — Computational Corroboration** Domain-general models trained only on raw data fail moral-reasoning benchmarks involving S unless supplied with an explicit “no unjustified killing” prior, confirming that data exposure alone is insufficient.
5. **P5 — Scenario Classification** In S, Person A’s killing is (i) intentional, (ii) unprovoked, (iii) yields no greater good, and (iv) occurs against A’s will. Therefore S satisfies the condition “unprovoked lethal harm” specified in P1-P4.
6. **P6 — Principle of Moral Evaluation ***(bridging premise)*
> **If** an act falls under a rule that the universal moral grammar forbids **then** that act is objectively wrong.
7. **C — Verdict** S falls under a universally forbidden rule (P5) and the principle in P6 applies ⇒ **S is objectively morally wrong.**