This study is one of forty-nine in the series "The Time of Questions and Answers", published on Meridian, the decision-record platform of the RISCON family. Every book in the series asks, within its own domain, what must be decided now and what would make the decision change. This one reads risk: what is taken, held or stopped before anyone knows how it ends, and the time between a question and its answer.
The making of it should be stated plainly. The design of the study — its question, its selection of cases, the works it may draw on and the reading notes that fix what may be said about them, and the plan the chapters follow — was made by an editorial session. The chapters were drafted by a model and checked by others. The evidence about markets and organisations in this study comes from one source only: the published decision records of WSTiger, a studio in the RISCON family that connects economic history and market signals to judge what risk to take now. Every block quote in the study is a record's own sentence, copied character for character, with its source named beneath it. Every record is a synthetic setting, a scene made for study. None of them reports a real firm, a real trade or a real market event. Where this study draws on a work of economics, psychology, history or organisation studies, it names the author and the work and stays within what is settled about it; where a reading is the study's own rather than the record's or the author's, it says so.
Why read decision records about risk at all. Histories written after the fact know how the story ended, and that knowledge quietly rewrites the reasons that stood before it. A decision record holds the other moment, the one before anyone knows: the setting, the question put as a choice, the options, a judgment offered as an example rather than a verdict, and a line saying when that judgment should be reconsidered. Because the record holds that moment, it can be read for what a decision writes down in advance and what it leaves unwritten — who decides, on what grounds and with what missing, by when, who waits and pays while the answer is not in, whether the decision can be undone, and what would make it change. That last pair is the subject of the series, and of this study.
The reader will find a literature review and a method before the cases, the cases read on six axes, and the hypotheses tested by counts. The study recommends no option. For each alternative in a record it shows what the alternative gains, what it risks, who is close to the cost and when, whether it can be undone, and what the record itself says would make its example judgment be reassessed. It says nothing about real markets.
The Introduction sets out the question and everything needed to hold it. It states the research question — how the time before an answer, the cost of waiting and the conditions for changing course are written down in WSTiger's records of risk decisions, and what that shows beside the main accounts of deciding under risk and uncertainty — and the four sub-questions that carry it: who decides and on what grounds; what the records say about deadlines, waiting and who bears it; whether decisions can be undone and what would make them change; and whether the same setting reads differently in the past, the present and the future. It names three contributions, each narrower than the work it builds on, describes the records themselves, says how the 53 cases were chosen by topic rather than sampled, sketches the method, and sets out what the evidence cannot show.
Chapter 1 maps the four literatures the study reads the records against: risk and uncertainty; organisations and accidents; waiting and irreversibility; and the judging of decisions and their makers. It begins with the background all four assume or contest — how probability became a way of deciding — and closes by saying where this study places itself among them. Chapter 2 sets out the method. It describes what a record holds, the six axes on which each record is read and the three further values beside them, the three kinds of cell in the ledger — copied, counted, and read by two models of different families — the strict and lenient readings, the eight hypotheses written down before any chapter was drafted, the rule that keeps a studio's writing habits out of the evidence, and the rule that what the record form guarantees is not a finding. It then codes two records in full.
Chapter 3 opens the reading of the cases with Knight's separation of the measurable from the unmeasurable uncertainty, Keynes on probability and the weight of evidence, and Hume on inference from experience beneath both. It reads four records at the edge of the measurable, and sets the debate over what Knight meant — LeRoy and Singell against Langlois and Cosgel — beside Ellsberg's ambiguity aversion and Kay and King's radical uncertainty. Chapter 4 takes the same problem into money markets: Bagehot's rule in a panic, Kindleberger's sequence of crises, Minsky's instability and Mackay's manias, against Lo's adaptive markets, read through four records of leverage, margin and collateral.
Chapter 5 turns to organisations: Turner's incubation, Vaughan's normalisation of deviance, Reason's latent conditions, and the argument between Perrow's normal accidents and the high-reliability school as Sagan tested it. It reads warnings that have sounded more than once and a stop button nobody has rehearsed. Chapter 6 reads shared measures — MacKenzie's performativity, Daníelsson's endogenous risk, Bookstaber's complexity, and Long-Term Capital Management as Lowenstein tells it — through four records of one measure used by many. Chapter 7 reads five records on judging a decision before and after its answer, with Tversky and Kahneman on heuristics, Kahneman's prospect theory on loss, Fischhoff on hindsight, Baron and Hershey on outcome bias, and Brier and Tetlock on what counted forecasts can test.
Chapter 8 begins the comparison across all the records, axis by axis: who decides and where the human's place sits when automated agents propose and commit, read with Simon on decision premises and Cohen, March and Olsen on organisational choice. Chapter 9 reads authority that travels by key, copy or instruction — six records from a two-key vault to a voice on a channel — with Saltzer and Schroeder's least privilege and separation of privilege as the borrowed design principles. Chapter 10 takes up the cost of waiting and who pays it: selling the tail in a marine insurance book and in weekly option premium, averages against expectancy, rewards paid now and losses borne later, read with Rajan, Taleb, Jensen and Meckling, and Luhmann's risk and danger.
Chapter 11 asks what can be undone: sizing and ruin with Kelly's fraction, positions that wait, the step from simulation to live trading, and the option value of waiting in Arrow and Fisher and Dixit and Pindyck, set against Collingridge's corrigibility. Chapter 12 reads the clocks inside the records — resets at midnight, trading days, counts of losses, expiries — and returns to settings the earlier chapters read in two or three times. Chapter 13 reads only the Invalidation lines: what every record states by its form, which conditions name something a reader could observe, and which leave matters to judgment, beside Clarke's fantasy documents and Neustadt and May's thinking in time.
Chapter 14 reports the ledger's verdicts on all eight hypotheses, under both readings, with their counts: which are supported, which the records go against, and which this corpus cannot judge. Chapter 15 sets those verdicts beside the earlier theories — Knight's distinction, the accidents-and-reliability debate, waiting and irreversibility, Luhmann and Power and Beck — and says again, narrowly, what the study adds and what it does not. The Conclusion answers the research question and its four sub-questions, restates the contributions and the limits, points to further research, and writes the study's own conditions for revisiting in words a reader could observe.
The counts the chapters use come from a decision ledger computed by code from the records, with two models of different families as coders. The corpus holds 53 records: 35 set in the present, 9 in the past, 9 in the future. They carry 159 options between them. All 53 records state a condition for revisiting the judgment, 53 conditions in all, of which 5 are stated with a number, a threshold or a time, which is what the ledger's code counts as observable. No record has a deadline field, and no record has a field naming the decider. On the cells the coders had to read, agreement varies: they settled who decides in 33 of 53 cases and who bears the waiting in 24 of 53, and they agreed far less often on whether a decision can be undone. Each count is given in the study with what it counts, and none is turned into a claim about a population.
Three limits should be held beside everything that follows. The records are synthetic settings with no outcomes, so nothing in this study can be checked against what happened. They were selected by topic, so "most records" means most of these 53 and not more. The coders are AI models whose agreement is reported rather than repaired, and the study itself is AI's expansion of the records, with its readings of the works kept within their reading notes. What the study offers is therefore a way of reading a decision before its answer — and of noticing what such a decision tends to leave out — and nothing more than that.