3.1The primary text
The primary text is David Hume's An Enquiry Concerning Human Understanding (1748), read in the Project Gutenberg transcription listed in the parent study's texts.json (https://www.gutenberg.org/cache/epub/9662/pg9662.txt). The file was downloaded for this article and its SHA-256 digest was checked against the prefix recorded in texts.json (024eb4c26ddd0993). The transcription is not marked as uncorrected optical character recognition, and every quotation from Hume below is copied from it. The transcription numbers Hume's paragraphs continuously, and quotations are cited by section, part and paragraph number as the file gives them. The file marks Hume's italics with underscores; quotations are chosen so that no underscored word falls inside them, since the format of this archive does not use italics.
The passages read are from Section IV, which Hume titled Sceptical Doubts Concerning the Operations of the Understanding: paragraphs 23 to 27 of Part I, and paragraphs 28 to 33 of Part II. Passages from Part II that the companion article quoted at length, including the bread passage of paragraph 29 and most of paragraph 32, are left aside or used only in short pieces, so that the reading rests on different ground.
3.2The records
The two records are published by WSTiger on its public atlas. Each sets a scene, poses a question with three options, gives an example judgment that WSTiger marks as its own dated reading, and adds an Invalidation line saying when the judgment would be reassessed. They are synthetic, modern summaries written for the platform, and not archival documents. They report no real firm, desk or market event. The originals are in Korean; the parent study holds an English translation of each, made once for the study and stored with a hash of the Korean original. This article quotes the English translation.
- Simulation to limited live: the record titled "From Simulated Operations to Limited Actual Trading" (wst:wt-atlas-expansion-b02-p26-q), dated "2026 · Synthetic Scenario". A trading system that has run stably in simulation is to be put to work on real capital.
- Split entry or averaging: the record titled "Split Entry and Loss-Averaging" (wst:wt-atlas-expansion-20260916-p10-q), dated "2026 · Synthetic Scenario". A trader whose position has lost wants to add to it, with no plan for the additional orders fixed beforehand.
Each record's Scenario field is laid out as an assumption, a list of the evidence the judgment requires, and tags. Each Invalidation field has two parts: a condition on which the judgment would change, and a line naming what is to be checked next.
3.3The ledger
The parent study reads 53 of WSTiger's records through a decision ledger. In plain words, the ledger is a table with one row per record and one column for each question the study asks of every record: who decides, on what grounds and with what missing, by when and with whom waiting meanwhile, whether the decision can be undone, what would make it change, and when it was judged. Some cells are copied from the record by code. Some are counted by code, such as whether a revisit condition contains a number, a threshold or a time. The rest are read by two models of different families, each giving a value, the basis for it (stated in the record's words, or implied by the setting) and a reason. A cell is settled when the two agree. The study registered eight hypotheses in advance, each with a rule for when the counts support it or go against it, read strictly (counting only what both coders found stated) and leniently (counting also what the setting implies).
Of the 53 records, 35 are set in the present, 9 in the past and 9 in the future, and all 53 are synthetic. Every record carries a condition for revisiting its judgment, because the record form asks for one; 5 of 53 (9%) of those conditions contain a number, a threshold or a time that code can detect. None of the 53 names a deadline, none names who decides, and none lists what is missing. These are features of WSTiger's record form as much as findings about any decision.
The ledger divides the question of undoing in two. One cell asks how hard it is to undo the restriction a judgment imposes (the reversal constraint). The other asks how hard it is to undo what the decision lets happen or stops (the reversal action). Both cells were hard for the coders. On the reversal constraint they agreed for 11 of 53 records (21%). On the reversal action they agreed for 13 of 53 (25%). These are the lowest agreement figures among the cells the study reports. The hypothesis that bears most directly on this article, H4, says that the harder a decision is to undo, the further apart are those who decide and those who bear the cost. Read strictly, it applies to none of the 53 records. Read leniently, it goes against: supports 1, against 3, unclear 3, not applicable 46, a share of 25%. Section 7 returns to what these figures can and cannot carry. In the case sections, the ledger appears only in a sentence or two of plain description; its cells are set out in the appendix.
3.4How the reading proceeds
Each case section opens with a claim about what Hume's text shows, reads that text at length, and then brings in a record. The aim is to see whether the record's own words take a position on the question Hume's paragraphs raise: what can be known about the effects of a step before it is taken, and what a small step can and cannot teach about a larger one. Each section states at its start and its end what it adds to the debate laid out in Section 2.