Conclusion

8.1Contribution

This article has argued that Sections IV and V of Hume's Enquiry state the claim that both sides of the organisational accident debate presuppose. Section IV shows that a run of uneventful experience gives no argument for expecting the run to continue. Section V shows that custom carries the mind across anyway, with the force of instinct, and that no reasoning can prevent it. Placed beneath the debate, Hume's argument explains why the normalisation of deviance needs no error of reasoning, and why any practice that resists it must work by changing the facts present to the people who decide, not by persuading them.

Read with the Enquiry beside them, the two synthetic WSTiger records each draw Hume's distinction between a calm record and a safe future, and each proposes a measure that would make the custom observable: the count of ignored warnings and the margins at which they sounded, and the worst case recalculated apart from the outcome. Their revisit lines name what would relax the caution, not what would end the calm. The question the article set was whether the absence of any accident so far rests on the custom Hume described and whether the records write what would break it. On the first half, the reading supports the identification, within the limits stated below. On the second half, these two records write the means of seeing the custom and leave the event that would break it unnamed.

8.2Limits

The limits are substantial and should be stated plainly. The records are synthetic scenes written for a platform; they have no outcomes and report no real organisation. They were not written with Hume in mind, and the correspondence this article draws is an interpretation. Two records cannot show a pattern; the ledger figures quoted in the discussion describe WSTiger's record form and 53 purposively chosen records, not organisations or markets. The English translation of the desk record departs from its Korean original at a point that matters, and other translation errors may remain unnoticed. The secondary works are drawn on only as far as the parent study's reading notes allow, which means that the positions of Vaughan, Turner, Sagan, Weick and the others are stated here in compressed form and may lose distinctions their full books make. Hume is read in a single public transcription.

Hume's own argument also limits what can be said. If custom is not open to reasoning, then this article's reasoning about custom does not show that any arrangement will succeed against it. The high-reliability school's claim that practice can hold the drift off is an empirical claim, and nothing here tests it.

8.3Further research

Three lines of work follow. First, the asymmetry in revisit lines could be examined across the parent study's other records, to see whether revisit conditions generally name what would relax a cautious judgment and rarely name what would show a calm had ended. Such a study would need care to separate what the record form produces from what the scenes contain. Second, the reading could be tried on real decision records where they exist and can be read, such as incident logs and limit-change approvals, asking the same question: when a limit is raised after a profitable breach, what fact does the approval document cite, and is it a run or a single present fact? Third, the high-reliability school's practices could be catalogued by which of Hume's mechanisms each acts on: changing what is present to memory or the senses, changing who decides, or changing when the decision is made relative to the outcome. That catalogue would give the debate a common vocabulary drawn from the argument both sides already assume.

Appendix: ledger figures used

All figures are from the parent study's numbers.md, computed by code from its decision ledger (2026-10-05), and from the ledger entries for the two records. The two coders are two models of different families.

Study-wide:

Coder agreement by cell across the 53: decider 33 of 53 (62%); bearer 24 of 53 (45%); reversal constraint 11 of 53 (21%); reversal action 13 of 53 (25%).

Cells for the two records: