The paper asked of sixteen authored AEGIS cases who must decide what by when, and what would make them revisit the decision. The answer the records give is consistent across the set. A single office, defined by the procedure at risk, owns each decision. The decision is due on the office's own operational clock, shortly before the next step would make a harm permanent or shortly after recovery, when an emergency measure would otherwise outlive its use. What is unknown at that moment is, in six of the sixteen cases, something only the applicant could say. And the decision is to be reopened when the harm it was meant to prevent becomes visible, watched for through an act of recording, but measured against no threshold and with no one named to bear the interval. Because all sixteen cases are authored synthetic settings, these are findings about how the AEGIS record set frames protective decisions, not about how any real office makes them.
Discussion
Set beside the literatures introduced at the start, the records take clear positions on some questions and are silent on others.
On where a protective decision belongs, they side with action close to the ground. Boin, 't Hart, Stern and Sundelius treat crisis management as a defining feature of governance in which leaders face critics looking for blame; Col describes local governments protecting citizens by taking the initiative while communities are on their own. The AEGIS records go further down than either: they place each decision with the one desk whose own procedure created the risk. This makes ownership unambiguous, and the case of the disbanded support team shows a record treating an unambiguous owner and a named deadline as the core of a commitment. It also means the records never ask what happens when the desk itself is the problem, a question the appeal-review case approaches when it asks how review can be independent in a small organisation, and then lists, among its three choices, review by another department or by an outside reviewer.
On warnings, the records turn the usual arrangement around. Lindell and Perry, and Mileti and Peek, study how people respond once a warning reaches them, and treat reception and understanding as the first steps of protection. In the AEGIS cases the office is the sender, and its protective decision is often about whether its own message arrived or was understood. The bounced notice and the rejection notice are, on this paper's reading, warning problems seen from the sender's side: the office must decide by a deadline whether to act on the chance that its message failed.
On precaution, the records keep the hold as one choice among three in a minority of cases and never let it frame the whole decision. Sunstein's case against the precautionary principle includes the observation that precautionary steps create dangers of their own. The records do not argue with that; they simply never price the hold. That silence is the point at which the burden literature becomes useful. Moynihan, Herd and Harvey argue that the costs of dealing with the state, and how they are shared between state and citizen, are often political choices. In several AEGIS scenes an applicant is the one in the interval, and no record says who bears the cost of the wait.
On revision, the records come closest to the incremental and adaptive views. Lindblom's administrator expects to repeat a limited comparison as conditions change; the pathways method of Haasnoot and co-authors commits to short-term actions and watches signposts to see whether adaptation is needed. Every AEGIS decision has a condition for coming back to it and something to watch. What is missing is the step that turns a signpost into a trigger: a number, a share, a duration. The backlog case has a calculated rate in its own record and still writes its reopening condition without one. And although every record asks the office to keep a record of what follows, every record also says its predictive performance was not measured, so the chance to learn that Kahneman and Klein name among the conditions for trusting a judgment is prepared but, on this paper's reading, not used.
Contributions
The paper makes three contributions. It gives a measured description of how one authored record set frames protective decisions under a deadline: owners lodged at the single responsible desk, deadlines on an operational clock rather than a calendar, unknowns placed with the applicant in six of sixteen cases, and reopening conditions written as visible failures without thresholds. It separates, by a stated count of repeated five-word runs, the parts of the records that carry case-specific content from those that repeat a template, finding the decision-bearing parts at 0% frame and the comparison links largely formulaic, so that later readers know which parts can carry an argument. And it places the set within the debate on deciding under deadline and uncertainty, showing that the records adopt the adaptive view that every decision should carry its way back while leaving unnamed what the precaution and burden literatures press hardest: who bears the time between a decision and its revision.
Limits
The limits are those of the material. The cases are authored synthetic settings; the study can say how they frame decisions, not how decisions are made. The selection is one rule applied to one programme of one studio; the other headings of the X100 set, and the thousands of other AEGIS records, may frame decisions differently, and the 2% share of "when" conditions across all authored records already suggests they do. The English is this study's own translation, made once and labelled; a different translator might have chosen other words, though the quotes can be traced to the Korean by the hash each English seed carries. The frame measure undercounts in Korean, which is why it was repeated in English within the sixteen cases; a measure on more than sixteen English records would be firmer. Finally, the coding of which unknowns concern the applicant was done by a stated rule on the Korean words for applicant, person, intention, delegation and residence; a different rule would draw the line in a slightly different place.
Further research
Three lines of work follow. The first is to apply the same questions to the other protective programmes in the AEGIS library, to see whether the operational clock and the desk-level owner hold beyond the X100 set or belong to it. The second is to compare the AEGIS framing with records from other studios in the RISCON family that carry deadlines and reopening conditions, using the same frame measure, to see whether thresholds are absent across the family or only here. The third is to ask, of any decision record, the question this set leaves open: who waits while the decision is watched, and for how long. A record that named the bearer of the interval alongside the owner and the deadline would close the gap this paper has described, and would let a later reader see not only who was supposed to decide, and by when, but who was asked to wait.