1.1Data and method
GAEKJU's public library held 496 questions when this study fetched it, every record read one by one through the studio's public interface on 2026-10-06. The library sorts them into three kinds, 40 record readings, 51 outside cases and 405 synthetic scenes, and into seven kinds of talk, each question asking one of three things of its kind of talk: whether it is the same across domains, whether it costs performance, and whether it has changed across eras. GAEKJU's English name for one kind of talk is "Missteps", and the library holds 56 synthetic scenes of that kind.
The selection follows one rule: every synthetic scene of the kind of talk "Missteps" whose evidence holds no digit. The evidence is the part of a scene that says what it rests on, what changes in its setting and what the scene leaves unanswered; a digit there marks a count, a rate or a year brought in beside the reasoning. The rule keeps the scenes whose evidence stands on words alone. It was applied by a script to the whole library before any scene was read, and it leaves 10 scenes and sets 46 aside. Five ask the domain question, three the era question and two the performance question. Their settings run from a generator works in the 1980s to a clinical-trial scheduling room supposed for 2030. None cites a published study.
Every scene is synthetic, and every one opens its evidence with the label "Synthetic:". The study reads each scene as a written frame for a decision, not as a report of what happens in a trade, and draws no fact from any of them about steelworks, gas inspection, hotels, clinical trials, assembly lines, fire lookouts, generator works, building sites, pharmacies or shops. Where a scene is read below, it is read as GAEKJU's construction.
The scenes come in two lengths. Six are short, with two lines of talk and readings of one sentence each. Four are longer, with three lines of talk and readings of two sentences. Both carry the same fields: setting, scene, talk, terms, evidence, three readings, a lean towards one of them, a judgment ending with the basis date 2026-09-25, a counter-condition and a proposal. These slots are guaranteed by GAEKJU's format, so the fact that every scene has three readings, a counter-condition and a proposal is not a finding. What the slots say is.
The records are Korean. This study made the English translation sentence by sentence, labelled in each scene's translated record with the hash of the Korean original; every block quote below is that translation, word for word. GAEKJU's Korean word for the kind, which means a stroke of the hand that misses, is put as "misstep"; the record's words for a wasted trip, an error and a confusion are kept apart in the English as they are in the Korean.
The counts in this chapter come from a script that reads only the fetched records. It tests the Korean for marks the argument needs: two records, marks, dates or appointments set against each other; judgments that decline to end in a person's fault; readings that take a slip as a discovery, a clue or a signal; the place where a correction is to land; proposals that divide the work after a slip; refusals to leave that work with one person. The tests are listed with their words beside the counts. Measured against the whole library, no field of the ten scenes repeats GAEKJU's own five-word runs to the series threshold of 30%; the highest, the judgments, stand at 2%. The scenes' sentences are their own, and can be read as such.
1.2A record decides
The first thing the scenes do is take the decision about what counts as a mistake away from the size of the loss and give it to a record. The plainest case is the scene of the first coil, set in a hot-rolling mill of the 1990s and synthetic. After the work rolls are changed, the first coil off the mill is downgraded until the thickness settles. The scene says why no one minds:
That loss is written in the production plan, so no one calls it a misstep.
The same week a second coil is downgraded when the roll gap slips, and a third-year member of the crew writes an incident report, saying in the talk that the first coil is the plan and this one an accident, though the same steel went into both. The scene does not settle whether the two losses are the same. One of its readings holds them apart and gives a reason for doing so:
The second coil is not the same thing as the first. A planned loss is the price of the rolls settling, but setting the gap wrong is an error of the hand, so putting the two in one slot blurs what can be learned from the error.
Another reading, the one the scene leans to, says what decides the name:
The name of a loss is set by the plan, not by its size. The first coil is not a mistake because it was written down in advance, and the third-year's coil is a mistake because that line was missing.
This is the scene's whole argument in two sentences, and it is a claim about records, not about hands. The judgment repeats it and draws one consequence:
The hot-rolling mill gives the same downgrade two names. What divided the names was not the size of the loss but whether the plan had that line. If there are days when the adjustment after a change is not all over by the first coil, part of the rolls' unsteadiness is passing into a person's incident report.
Read beside Reason, Human Error: Models and Management (2000), the scene sits on the line between the two approaches. The person approach would read the second coil as a slip of the crew member's hand; the system approach would look upstream for the trap. The scene does something narrower than either. It does not say the crew member was blameless, and the condition it gives for being wrong names the record that would show the second coil to be an error of the hand after all. What it says is that the label "mistake" was assigned by whether a line existed in a plan. The scene's second reading adds a point that Hollnagel (2018) would recognise: the incident report records only the one coil lost, not the two coils kept when the gap was caught. This study reads that line as a small instance of the ledger Hollnagel describes, which counts what fails and has no column for what goes right; the reading is this study's own.
The scene of the empty doorstep, set among city-gas inspectors in the 2020s and synthetic, moves the same question onto a wasted trip. The inspector's line declines the word "wrong" outright:
That's the third sticker. It's not that I got it wrong; my hours and this household's hours don't match.
Here the two records are the office timetable and the residents' day, and the third record, the one that turns their mismatch into a fault, is the monthly sheet that lists uninspected homes by inspector. One reading says so:
The three wasted trips are the price of the office timetable and the residents' day not matching. As that mismatch is written on the sheet of uninspected households by inspector, the price the schedule should pay moved onto one person's legs.
The scene does not let this reading stand alone. Another reading gives the residents their share, observing that some of the doors shut on weekday daytimes are the choice of residents who put off the inspection. The judgment then leans the other way without naming anyone at fault:
In the third wasted trip no one got anything wrong. Yet the price is paid only by the inspector's legs and the month-end sheet. That two homes opened through the Saturday number means the wasted trip was a signal telling of mismatched hours.
The two shorter scenes on this ground make the same move in fewer words. In the scene of the room desk, set at a hotel front desk and synthetic, a room change that worked the day before is blocked, and a colleague points out that "cleaning done" and "available for sale" are different marks. One reading puts the difference down to a confusion on the part of the person in charge; the judgment declines to stop there:
The lead's words are a clue for finding where things are blocked. Only by checking whether the cause is that 'cleaning done' and 'available for sale' were merged into one status does it not end as a person's fault.
In the scene of the two dates, set in a clinical-trial scheduling room supposed for 2030 and synthetic, the error is not in the copying at all:
The automatic entry was accurate, but reading the two appointments as the same thing is the problem.
The script finds that six of the ten scenes set two records, marks, dates, times or appointments against each other, and that five of the ten judgments decline to end in a person's fault or put off praise and blame. Four scenes keep at least one reading that names an error, a confusion or a fault of a person or a hand, so the person is not erased; the person stays as one reading among three. Vaughan, The Dark Side of Organizations (1999), finds routine nonconformity produced by the interconnection between environment, organisations, cognition and choice. The scenes stage a small and concrete version of that interconnection: a plan and a mill, a timetable and a day, a status screen and a room. This study reads the seam between two records as the place where, in those six scenes, a loss becomes a mistake; the reading is this study's, not Vaughan's.
1.3The slip as news
The second thing the scenes do is read some slips as information. The script finds five of the ten whose readings or judgment take a slip, a workaround or a wasted trip as a discovery, a clue, a signal or the price of keenness.
The scene of the makeshift rest, set on a surface-mount line in the 2020s and synthetic, is the clearest. The jig that holds a part does not fit a new part, and a technician carries on with a makeshift rest. The technician's line reports two things at once:
The rest didn't fit, so I used something else for a moment. The dimensions seem to have changed.
The first sentence is a workaround; the second is a finding. The scene's three readings split exactly on that double nature. One calls the rest a choice outside the specification and so risky; another says it reveals a difference in size the design missed; and the one the scene leans to refuses both:
Before it is verified, it cannot be settled as either a risk or a discovery.
The judgment holds the same line:
The technician's skill kept the day's work going and can show a difference the design missed. Whether it can be used in the product needs a separate check, so a conclusion praising or condemning the invention is put off.
The scene reads like a small case for the literature on workarounds. Debono, Greenfield, Travaglia, Long, Black, Johnson and Braithwaite, Nurses' Workarounds in Acute Healthcare Settings (2013), conclude that workarounds enable, yet potentially compromise, the work they serve; the scene's readings are those two halves, and its lean is the refusal to choose before a check. Alter, Theory of Workarounds (2014), proposes studying how workarounds sometimes lead to larger planned changes in systems; the line lead's suggestion that the designers see the difference is a scene of that route, from a makeshift fix to a changed drawing, written before anyone knows whether it will be taken. Frese and Keith, Action Errors, Error Management, and Learning in Organizations (2015), count innovations among the positive consequences error management seeks. The scene names the rest an invention, and then declines to praise it until it is checked. This study reads that refusal as a guard against reading a positive consequence into a slip too soon; the reading is its own.
The scene of the fire lookout, set at a forest-fire lookout in the 2010s and synthetic, turns the same question onto a false alarm, and it is the scene that weighs the cost of reading a slip as a fault. A lookout reports smoke that turns out to come from burning the ridges of a paddy, and a single "Again?" comes back over the radio. The next day the lookout watches ten more minutes before calling. The veteran at the next post answers on the radio that evening:
By the time you're sure, it's already a big fire. Even if you call ten false smokes, if you call the real one first, that's our job. I called in plastic being burned twice last week myself.
The reading the scene leans to sets the price of the reproach:
The number of false reports is another name for how keenly the lookout watches. If a single 'Again?' raised the threshold for reporting by ten minutes, words meant to cut missteps moved the price onto seeing a real fire late.
and the judgment names where that price lands:
The lookout's false report is closer to the price of keenness than to being wrong. If a word of reproach raises the threshold for reporting, that price moves to the most expensive place, the first ten minutes of a real fire. What guarded the eyes of this post was the veteran's talk, which brought up the veteran's own false reports first and set the threshold back.
The scene keeps the other side in view. Another reading holds that watching ten minutes more was judgment, not fear, since a call made after seeing the colour and spread of the smoke spares the crew wasted trips. And the scene states plainly what it cannot answer: how much the ten minutes' delay held back the discovery of an actual fire. The judgment, in other words, is a claim about how a false report should be named, and it leaves the size of the risk open.
Dixon, Wickens and McCarley, On the Independence of Compliance and Reliance (2007), found in an experiment that automation prone to false alarms hurt overall performance more than automation prone to misses. The fit with the scene is partial, and this study says so: their false alarms came from a machine and their measure was the operator's performance, while the scene's false alarm comes from a person and its question is what the person who raised it will do next time. What the two share is that a false alarm is found to carry a cost: the experiment measured it in the operator's overall performance, and the scene finds it in the reproach that follows. The scene puts the cost on the far side of the trade, in the late call; this reading of the fit is this study's own.
The scene of the buried notice, set on a building site seen in two eras and synthetic, offers a different kind of news: not a finding about the work but a change in when an error can be seen. In the 2000s a foreman who misheard a column number over the radio asked back at once; in the 2020s an order posted to a group chat of forty foremen is buried under photos and confirmations, and one foreman closes the formwork in the old order. The scene states what it does not know and what it suspects:
Whether errors actually increased on this site was not observed. An error in the radio days showed at the moment it was heard, and an error in the group-chat days shows after the formwork is closed, so what changed may be the time an error comes to light rather than the amount of error.
One reading then turns the chat into a record in the other sense:
The group chat did not increase errors; it left a record. Among the mishearings of the radio days there must have been many that passed without anyone asking back, and now at least who posted what and when remains.
The scene thus holds two opposite readings of the same channel, one in which it hides an error and one in which it keeps a trace, and its judgment chooses only the narrower claim, that the two errors are not the same error.
1.4Where the correction lands
The third thing the scenes do is ask where a correction has to go to hold, and who carries the work it takes. The script finds seven of the ten whose scene, readings, judgment or proposal name the record or place where the correction must land, and six that refuse, in a reading, the judgment or the proposal, to leave the work after the slip with one person.
The scene of the pencilled drawing, set at a generator works in the 1980s and synthetic, is the most compact. A foreman corrects a dimension on a paper drawing in pencil and fits the part. The colleague's line is both an offer and a rule:
I'll go with you to the person who will change the drawing. We can't leave the pencil to work alone.
The judgment names the place the correction has to reach:
The foreman's hand saved today's part. To save it in the next build too, the dimension has to reach not one person's memory but the place that holds the right to make corrections.
The scene separates two things that a count of fixed mistakes would merge: the part that was saved and the drawing that was not yet changed. Error management, in the sense van Dyck, Frese, Baer and Sonnentag, Organizational Error Management Culture and Its Impact on Performance (2005), give it, includes correcting errors quickly as a shared practice. The scene adds a condition that this study reads as its own contribution: a correction made quickly and in the right place may still be lost if it is made in a record that has no authority over the next build.
The proposals of the other scenes mostly make the same move on a different record. The first coil asks for one more free slot in the production plan, so that the unsteadiness after a roll change is borne by the plan before it passes into an incident report. The empty doorstep asks for the record of uninspected homes to be kept under the visiting hours rather than under inspectors' names, returning the price of wasted trips to the side of the schedule. The room desk asks that the front desk see the cleaning and sale statuses change together. In each case the correction is not an instruction to a person to try harder but a change in the record that names the next loss.
The second half of the move concerns who carries the work after the slip. The scene of the swapped container, set in a hospital pharmacy in the 2020s and synthetic, puts it as a condition on accepting a stopgap:
The gain of the stopgap is acknowledged, but the person who traces must not be made to link the information again alone later.
and its proposal says it again:
Divide dealing with the wait and checking the link among different times and people. Do not leave the aftermath of a hurried choice to one pharmacist.
The scene of the missed handover, set on a shop floor in the 2020s and synthetic, gives the same answer to an admitted omission. The manager is the first to say that a customer's message was missed, and writes it up at once. The judgment credits the admission with letting the next shift know, and then sets a limit on what the admission does by itself:
The manager admitting it first let the next shift know the situation. But the time needed to keep the promise to the customer has to be set aside separately.
and the proposal protects the person who spoke:
Schedule the time for explaining to the customer side by side with the shift handover. Do not pile all the later explaining on the manager who found the omission.
The buried notice makes the same point about the fix itself. The manager's offer to phone forty foremen restores the asking-back the radio once gave, and one reading warns where that labour would fall:
The manager's offer to phone each one brings the vanished asking-back in again by voice. If that labour is loaded onto the phones of one or two people in the site office, a few people end up paying for forty foremen's asking-back.
The script finds seven proposals that divide the work after the slip among times or people. Frese and Keith (2015) argue that learning from errors occurs under certain circumstances, one being a mind-set of acceptance of human error. The scenes write a more practical circumstance into their proposals: the one who finds, admits or fixes a slip is not left to carry everything that follows. This study reads that clause as the scenes' answer to a question the error management literature leaves in the background, namely whose time the management of an error takes; the reading is this study's own. Halbesleben, Wakefield and Wakefield, Work-arounds in Health Care Settings (2008), discuss how open discussion can let work-arounds improve work processes. On this study's reading, the swapped container shows the condition under which such openness would cost the one who opens it least: when the link the stopgap breaks is given its own time and its own person.
Three scenes stand outside the count of divided proposals: the first coil, the empty doorstep and the pencilled drawing. Each of them, as read above, moves the correction into a record instead, a slot in the plan, the visiting hours, the official drawing. The scenes thus offer two ways of keeping a slip from staying with one person: share the work after it, or change the record that names it.