review_job_id: 8382
review_pair_id: 20916
note_path: kb/notes/naur-binds-theory-to-humans-via-premise-that-machines-follow-rules.md
criterion_path: kb/instructions/critique-note.md
model_partition: codex
runner: /root/closing_critique
runner_model: null
runner_effort: null
result_kind: report
outcome: null
completed_at: '2026-08-26T14:46:34+00:00'
Critique: Naur's rule-inexpressibility argument does not by itself bind program theory to humans
Note: kb/notes/naur-binds-theory-to-humans-via-premise-that-machines-follow-rules.md Central commitment: The inexpressibility of program theory as rules cannot by itself establish that only humans can possess program theory; human exclusivity requires an additional restriction on eligible theory-holders. Critique mode: claim Attack outcome: no surviving attack
Strongest case against it
A Naurian conceptualist would reject the note's separation of inexpressibility from human possession as an anachronistic reconstruction. On this reading, Naur does not first specify a bearer-neutral capacity called “program theory” and then infer which species may possess it. He adopts Ryle's personal notion of theory: possession consists in intelligent performance together with the ability to explain and justify it, and program theory is the programmer's situated understanding of how affairs in the world relate to a program. A computer program, by contrast, is formal symbol manipulation. The regress is meant to show that the intelligent grasp constitutive of theory is not rule-following, while a machine qua computer program does only what formal rules determine. “Nonhuman machine theory-holder” is therefore a category mistake inside Naur's conceptual scheme, not an untested member of a bearer class. The human-binding conclusion follows from what theory-holder, programmer, and machine mean in the argument, so no independent empirical survey of possible nonhuman bearers is needed.
The linked source analysis gives this objection its strongest textual form. Its retained passages describe relevance as contributed by a programmer who understands the world, theory as the mental possession of each programmer, programming as matching the world to formal symbol manipulation, and similarity judgment as accessible to human theory-holders but outside determination by rules. Those statements can be read as one constitutive contrast rather than as a loose empirical generalization from known humans and machines.
How the note engages it
Fully answers. “Human exclusivity needs an additional premise” explicitly identifies both ways the objection can bind theory to humans: Naur's person-bound descriptions of theory and his human-judgment/rule-following-machine opposition. The note does not require that the bridge be an empirical survey; it allows Naur's definitions or conceptual boundary to supply it. Its point is instead about which claim does the inferential work. If “theory-holder” is stipulated to be a human programmer, that stipulation supplies human exclusivity rather than deriving it from rule-inexpressibility. If machines are stipulated to exhaust the rule-following pole, that machine characterization is the bridge premise. In either case, the conceptualist objection relocates the exclusion into a definition and thereby confirms, rather than refutes, the narrower claim that inexpressibility alone does not entail human-only possession.
“A trained interpreter exposes the question but does not settle it” also answers the objection's machine half without presuming that current LLMs possess theory. It distinguishes a human-readable rule set from formal determination and computability, grants that a formally implemented learned model remains on Naur's machine pole under a broad reading of rules, and concludes only that the essay leaves the relevant reading underspecified. The note therefore neither smuggles in an LLM counterexample nor assumes that a nonhuman theory-holder actually exists.
Constructive findings
- None — the note already answers the strongest objection. Its central commitment is the limited logical claim that the inexpressibility proposition is insufficient on its own, and the best conceptualist reading of Naur still needs the person-bound definition or machine-as-rule-follower characterization that the note identifies.
Secondary objections (optional)
- “Excludes every eligible nonhuman theory-holder” gives Naur's sentence a stronger modal scope than his practical contrast may bear. The linked source centers human programmers and computer programs; it does not show that Naur considered animals, extraterrestrials, or every possible artificial architecture. Framing the target as Naur's human-binding wording, rather than as a universal taxonomy Naur explicitly argued, would make the exegetical scope more precise. This does not affect the logical claim about inexpressibility alone.
- The three transfer tests are plausible constraints on an interpreter-plus-artifact system, but they are not all established by the compiler case. In particular, that case supports one failed acquisition route and repeated control by the original team; it does not independently demonstrate the stated dispositional-reliability threshold for every theory-holder. This is a secondary extrapolation, not a defect in the note's central commitment.