Conclusion

8.1Contribution

This article has connected the option-value argument for waiting (Arrow and Fisher 1974; Dixit and Pindyck 1994), Collingridge's dilemma of control (Collingridge 1980) and Kelly's arithmetic of sizing (Kelly 1956) through Section IV of Hume's Enquiry. The connection is that the information whose later arrival gives waiting its value is, for some steps, information that only the step can produce, because the effects of a new object are learned only from experience of it, and experience of a similar object informs only about itself. Where that holds, the real choice is between committing in full and committing in part. Hume's separation of law from application in paragraph 27 adds that an exact sizing rule rests on an edge that only experience estimates.

Two synthetic records written for a decision platform were read against this. Both divide the irreversible step. One buys experience with a restricted, approved and stoppable exposure; the other sets a plan and a full calculation of loss in front of an addition. One names the right kind of experience to judge the stage by, operations indicators rather than early profit. The other refuses to let a plan made after a loss govern the position that lost.

8.2Limits

The records are synthetic, modern summaries written for a platform, not archival documents. They report no real firm or trade and have no outcomes, so nothing here shows whether staging works. They are translations of Korean originals. The record form requires a revisit line and has no field for a deadline or a decider, so the absence of those is partly a feature of the form. The ledger's coders are models, and their agreement on reversibility is low. The readings of the secondary works go no further than the parent study's notes on them, and their placement in the debate is this article's own. Hume wrote about inference, not markets.

8.3Further research

Three lines follow. The first is to read records whose settings have outcomes, so that a staged decision can be compared with what happened next. The second is to sort a larger set of records by whether the information they need arrives without the step, and see whether the records' choices follow that line. The third is to ask, of any staged decision, who approves the passage from one stage to the next and on what number. Neither record answers that question.

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:

Cells for the two records: