1.1Data and method
The records are NADIR's public list of question–judgment pairs, fetched once by GET on 6 October 2026 (UTC) and kept in this study's folder. The list holds 1534 pairs. Every one has the status "hypothesis", none carries axes or evidence keys, and the list's evidence table is empty. The studio files each pair under one of 10 layers and one of three phases of its research frame: erosion (2016 to 2025), collapse (2026) and transition (2027). Its own description of the batch speaks of 992 Korean research drafts, and the list labels the authorship of each pair. Four labels occur. Three of them, the hypothetical decision exercises, the episode-derived interpretations and a smaller group of blackbox interpretations, come with options; 1038 pairs in all set out options to weigh. The fourth label, "NADIR episode interpretation draft", is carried by 496 pairs, and these are exactly the pairs with no options: every draft lacks them, and no pair outside the drafts does. The drafts are spread across the frame by horizon. There are 267 in the erosion phase, all with the horizon "past"; 111 in the collapse phase, all with the horizon "present"; and 118 in the transition phase, all with the horizon "future". In the collapse phase every one of the 10 layers has between 11 and 12 drafts. All 496 are unreviewed.
The selection rule is one sentence. This paper reads every pair in the layer "learning", which the studio labels memory and learning, in the phase "collapse", whose authorship reads "NADIR episode interpretation draft". The rule gives eleven pairs. The layer holds 187 pairs, 20 of them in the collapse phase; the 9 collapse-phase pairs left out are the ones that set out options. The 27 learning drafts in the erosion phase and the 11 in the transition phase are not read, so that the pairs read share one moment of the frame, the year of the collapse, and one horizon, the present.
Each pair read is short. It has a question and a judgment of exactly two sentences, and the eleven questions and judgments together come to 552 words in this study's English. Ten of the questions are put as questions and one as a noun phrase. Ten pairs belong to the studio's fourth edition and one to its third. None has options, axes, evidence keys or a stated time at which its judgment became available. The studio does not say which episode, if any, lies behind each draft, and the drafts name no real person, organisation, place or event; this study treats each as an unreviewed interpretation written on the research frame, and nothing in the appendix had to be withheld for naming one.
The records are in Korean. This study's editorial session put every string into English, sentence by sentence and with nothing added, and the Korean stays the original. Every block quote in this paper is a sentence of that English, word for word, with its source line naming the pair by a short title (listed in the sources table) and the field: Question or Judgment. Every count in the paper comes from a script that reads only the kept records and prints its results to a table in the study's folder. The counts of wording, such as how many judgments ask for a distinction, were made on the Korean originals by fixed word lists that the table prints beside each count.
Two cautions from the series' rules apply. A field whose wording repeats across many records belongs to the studio's frame and is not evidence about any one record. Among the eleven pairs no run of five words recurs in three or more questions or three or more judgments, so the wording read here is each pair's own. And a property that the studio guarantees to every pair is not a finding. The drafts carry no options and no slot for reopening a judgment, so this paper's finding about time, in section 1.3, is about what the judgments say in their own words, not about a slot left empty.
Finally, what the paper does not do. It does not test whether the drafts are true of any organisation, since nothing in them was observed and the studio itself calls them conditional proposals. It does not recommend any way of keeping records. It reads how a published set of interpretations frames one kind of decision.
1.2What the knowledge of the ending does to the record
The drafts' first concern is the reader who already knows how the story ended. The one draft from the studio's third edition states it most fully. Its question is a noun phrase, not a question, and the phrase names a loss:
The other paths that vanish when read again with eyes that know the ending
The judgment then says what is lost and what must be done about it:
The outcome of the ending makes the events that led to failure easy to remember and tends to erase the events that showed a possibility of recovery. To make ten years of experience into knowledge, the actual path and the other paths that were possible at the time must be kept together.
Read closely, the judgment does three things. Its first sentence is a claim about memory, not about the decline. The ending is the subject of the sentence and the agent of the forgetting: it is the outcome that makes some events easy to remember and erases others. Its second sentence turns a claim about memory into a requirement on a record. Experience becomes knowledge only if the record holds two things at once, the path taken and the paths that were open. And the two are tied to a time: the other paths are those "possible at the time", not those that look possible now.
That requirement is close to what Roese and Vohs, Hindsight Bias (2012), report about the bias: knowledge of the outcome narrows attention to a single causal account and neglects other reasonable explanations, and asking people to consider other explanations can reduce the effect. This paper reads the draft as moving that remedy from a person to a record. Roese and Vohs describe something that happens in a mind. The draft asks a document to carry the alternatives so that the mind that later reads it does not have to reconstruct them. Whether a record can do that is not something the draft, or this paper, can show. The draft only frames it as the thing to be decided.
The same concern returns in a sharper form when the draft asks about the last event:
If only the last failure is analysed, can ten years of decline be understood?
The ending event happens on top of resources and relationships that had already shrunk. The last shock must be distinguished from the cumulative process that made that shock impossible to withstand.
Here the work of the judgment is a distinction. The ending event is not denied; it is placed. It happens "on top of" something, and the record must tell the two layers apart. This is the first instance of the form that most of the drafts take. Of the eleven judgments, 4 ask for two things to be distinguished or assessed separately and 4 ask for two things to be kept or looked at together; 8 do one or the other. The decision posed is not between two courses. It is between two ways of filing the same events.
A third draft names the error that hindsight invites when the outcome was extreme:
Is an extreme outcome hiding causes that were ordinary?
It is easy to expect a large failure to have a special cause.
The answer it gives is again a rule of weighting for the record, not a course of action:
The path along which small concessions and omissions were repeated must also be examined with the same weight.
Two features of the wording are worth noticing. The first sentence describes an expectation as "easy", a tendency rather than an error. Six of the eleven judgments describe what can, may or tends to happen before they say what must be done; the drafts present the distortions of memory as ordinary, not as faults of particular people. The second sentence uses "must", as 7 of the judgments do. The obligation falls on the examination, not on any named examiner.
Two drafts turn hindsight onto the people inside the failure. One asks about the warner:
May the interpretation of the person who warned first be taken as right in every part?
Pointing accurately to some risks does not verify every explanation of the cause. The warning's hitting the mark and the soundness of the solution that person proposed are assessed separately.
This is the only judgment of the eleven whose closing sentence is a plain statement in the present tense rather than a "must" or a "needed". It does not say the warning and the remedy should be assessed separately; it says they are. The effect is to make the separation sound like a procedure already in place. What the draft guards against is a second kind of hindsight. Once the failure has happened, the person who saw it coming gains a standing that extends to everything that person said. The draft splits that standing into two parts and keeps only the first.
The other draft concerns a judgment that changed while the crisis was under way:
Can a judgment changed during a crisis still be explained later?
Urgent decisions tend to go ahead without a record, so afterwards they can look like caprice. It is important to leave, even briefly, the facts newly learned at the time and the alternatives given up.
This is where the drafts come closest to the moment of the decision itself. The danger named is that the reasons for a change of course are never written, so that a later reader, who knows the outcome, sees only inconsistency. The remedy is two things to be written down at the time: the new facts and the alternatives abandoned. It is the same pair that the third-edition draft asked for, the path taken and the paths open, now at the scale of a single decision. Read through Roese and Vohs, this paper takes the draft to be asking for the record that would let a later reader resist a hindsight judgment of the decision maker, not only of the events. The draft says "even briefly". It asks for the minimum that would make a later judgment fair.
A fourth draft in this group puts the same worry the other way round, as a pressure on belief during the crisis itself:
Is a simple explanation of the cause believed all the more as events grow more pressing?
Whether clarity in a crisis is a simplification of the facts or a priority for action must be told apart.
Here the distinction is between two meanings of one quality. Clarity in a crisis may be a way of acting, deciding what comes first, or a way of believing, cutting the facts down to what is easy to hold. The draft does not condemn either. It asks that the record say which it was. The decision, once more, is a decision about how events will be classified.
Taken together, these drafts frame the year of collapse as the year in which the record of the decade is most at risk, and they frame the risk as the one Roese and Vohs describe: a single causal account that knowledge of the outcome makes look inevitable. Their answer is not a course of action but a demand on the record to hold pairs: the last shock and the slow erosion, the warning and the remedy, the decision and the alternatives given up. What the drafts add to the hindsight literature, on this reading, is a shift of the remedy from the reader's mind to the writer's file.
1.3Who writes the record, and who is left out of it
The second group of drafts moves from what the record contains to who makes it. Only one question of the eleven asks "who", and it asks about the selection of the record itself:
Who should choose the materials to be left for afterwards?
If the party responsible alone chooses the materials to be kept, there is a risk that a memory favourable to that party is what remains. A perspective is needed that examines the criteria by which the record is selected and whether anything has been left out.
The answer is worth reading slowly, because it answers a different question from the one asked. The question asks who should choose. The judgment says who should not choose alone, the party responsible, and then, where a person might be expected, it names a "perspective". It does not say whose perspective, or who would hold it, or what authority it would have. The record is to be examined for its criteria and its omissions, but the examiner is left unnamed.
This is close to what Brown, Stacey and Nandhakumar, Making Sense of Sensemaking Narratives (2008), describe in another setting: people tell the story of shared work in ways that serve the selves they are constructing, through impression management and attributional egotism. The draft frames the same tendency as a property of an archive. The danger is not that the responsible party lies; it is that a memory favourable to that party "is what remains". The verb belongs to the store picture of memory: something is kept and something is not. The worry, though, is about practice, about whose interest shapes what is kept. In the terms of the critique by Rowlinson and others, Social Remembering and Organizational Memory (2010), the draft speaks of a storage bin and fears a social process. This paper reads that double vocabulary as the most characteristic feature of the set.
The next draft names a second way in which a record loses its value, by being made easy to read:
What is lost when the record of a failure is tidied too smoothly?
If contradictions and uncertainty are erased, why the choices of the time were hard can no longer be seen. Conclusions that are easy to read and a record that preserves the actual process of judgment need a balance.
The first sentence links the hindsight drafts to this group. What is erased, contradictions and uncertainty, is exactly what knowledge of the ending removes; what is lost is the difficulty of the choices "of the time". The second sentence then declines to choose. It does not say the readable conclusion should give way to the full record, or the reverse. It says the two "need a balance". This is the clearest instance of a judgment without options. A draft with options could have offered a readable summary against a complete file, each with a cost. The draft keeps both and leaves the weighting to whoever keeps the record, whom it does not name.
The sharpest draft in the set concerns silence:
May all the silence of the time be recorded as consent?
In an unfavourable situation people may stay silent because they have neither the room nor the influence to speak. When a record is interpreted, the positions that were expressed and the conditions under which they could have been expressed must be looked at together.
The question is put in the language of record-keeping: may silence be "recorded" as consent. It is not about whether people should have spoken. It is about how their not speaking should be written. The first sentence of the judgment gives two reasons for silence, the lack of room and the lack of influence, and neither is agreement. The second sentence asks for another pair to be held together: what was said, and the conditions under which it could have been said.
Milliken, Morrison and Hewlin, An Exploratory Study of Employee Silence (2003), found that the reason most often given for staying silent was fear of being labelled negatively and of damaging relationships, and they model how the social consequences of speaking up can take away a person's influence. Morrison, Employee Voice and Silence (2014), describes silence as depriving the organisation of potentially useful information. The draft carries that loss forward in time. On this paper's reading, information withheld in the meeting is lost once; a silence then recorded as consent is lost a second time, because the record would then say the opposite of what the draft supposes happened. Trouillot's two kinds of enforced silence, the unconscious and the deliberate, give this paper its reading of the draft. This paper reads the draft as asking the record keeper to treat each recorded silence as a possible instance of the first kind, and to keep the conditions that would let a later reader tell.
A fourth draft widens the point from silence in a room to absence from the whole account:
Does the memory of those who suffered less harm become the centre of the record?
Those who have time and resources left may get more chances to explain the events. For the record to be representative, the experiences of those who had no room to speak must also be sought out.
The mechanism named is not malice but capacity. Those who explain the events afterwards are those who have time and resources left after them, and in the draft's picture of a collapse those are the people least hurt by it. The record's centre of gravity drifts toward them without anyone deciding that it should. The remedy is the only one in the set that asks for an active search: the experiences of those without room to speak "must also be sought out". This paper reads the draft beside the forgetting work that Mena and others, On the Forgetting of Corporate Irresponsibility (2016), describe, which includes the silencing of vocal rememberers. The draft describes a quieter version of the same outcome. No forgetting work is needed if the rememberers never had the room to remember aloud. The draft names the drift; it does not name who would do the seeking.
The last draft turns from the record to what the record is for:
Is knowing why it ended the same as being able to begin again?
Even when the causes are understood, there may be no capacity to build resources and relationships anew. The completeness of an explanation and readiness for the next action must be seen as different stages.
Every other draft in the set asks that the record be made more complete. This one places a limit on what completeness can do. A full explanation of the failure is one stage; the capacity to act again is another, and the first does not supply the second. It is the only pair of the eleven that names a next action at all. Read beside Casey and Olivera, Reflections on Organizational Memory and Forgetting (2011), who ask for theory on the role of time in memory, the draft separates two times that the other drafts fold together: the time of understanding and the time of acting.
Two features of this group, read as a whole, bear on the series' question of what would make a decision change. First, the drafts name people only as sources of distortion or as the distorted: the party responsible, the person who warned first, those who suffered less, those with no room to speak. Three pairs name a person, a party or a perspective in so many words. None names a person or body who would keep, examine or correct the record. Where a keeper is needed, the drafts supply a "perspective" or a passive verb. Second, the drafts give no time. No pair names a date, a deadline or a moment to look again. Two pairs speak of "later" or "afterwards", and 4 speak of "the time", meaning the time of the events. The record is to be made now, in the year of collapse, and read at some unstated point after it.
The result is a particular form of the condition for changing a judgment. A judgment with options can say what signal would reopen it. These judgments say what a reopening would need to find in the record: the other paths, the slow process, the abandoned alternatives, the conditions of silence, the unheard experience. The condition for judging again is kept as material in an archive, not set as a trigger. That is the sense in which the set frames the decision about memory as a decision about a later reader.