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 "Holding back", and the library holds 59 synthetic scenes of that kind.
The selection follows one rule: every synthetic scene of the kind of talk "Holding back" whose evidence and three readings hold no digit. A digit in those fields marks a count, a rate or a year set beside the reasoning; the rule keeps the scenes whose reasoning stands on the written scene alone. It was applied by a script to the whole library before any scene was read, and it leaves 13 scenes. Five ask the domain question, four the performance question and four the era question. Their settings run from a women divers' fishing ground in the 1970s to a pest-control team supposed for 2030. None cites a published study.
Every scene is synthetic. The study reads each one 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 demolition, schools, laboratories, couriers or farms. Where a scene is read below, it is read as GAEKJU's construction.
The scenes come in two forms. Five open their evidence with the label "Synthetic:" and are short, with two lines of talk each and three readings of a sentence apiece. The other eight are longer, with three or four lines of talk and readings of two sentences or more. Both forms 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. Because some of these slots are guaranteed by GAEKJU's format, 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 name for the kind, the body that pulls itself out, is kept in the translation as "the body that holds back".
The counts in this chapter come from a script that reads only the fetched records. It tests the Korean for marks the argument needs: words for a load, a share, a burden or a price; the persons named as receiving it; the words with which a judgment sets a motive aside; a way of paying or counting named in the scene; a personal motive left standing in the counter-condition. The tests are listed with their words beside the counts. Measured against the whole library, no field of the thirteen 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.
1.2The one who takes it on
The first thing the scenes do is give the work that was put down a destination with a voice. In the repair counter scene, set in the 1990s and synthetic, the clerk who sends customers to the workshop says only that the customers keep arguing, and the people who answer are the chief who comes out and the young mechanic left at the bench:
While you're out, chief, two tillers are waiting. I'll drain the oil first.
The line is small, but it shows where the scene looks. The question in its title is why the bad news passes over to the mechanic, and the evidence names what passed:
Here what passes over is less the amount of work than the emotional load of taking the customer's disappointment.
This is the move the scenes make again and again. Holding back is written from the side that receives it. The script finds that twelve of the thirteen scenes speak, in their readings or judgment, of a load, a share, a burden, a cost or a price, and ten name who receives it: a colleague, the next shift, the person who asked for help, a customer, a newcomer, waiting farmers, the deep divers.
The complaints counter scene, synthetic and short, puts the matter in its plainest form. A team leader leaves the counter for twenty minutes to set out why the same complaints keep coming back, and the clerk who stays takes the queue. The clerk's line is both an agreement and an invoice:
I'll take the queue. Next time please schedule the time for setting it out as well.
The judgment weighs two goods and refuses to fold them into one person:
The gain of the setting-out and the cost of the waiting should not be tied to the same person's character.
Here the scene does what Daniels, in Invisible Work (1987), asked of the concept of work: it brings into view a labour that a plain account would miss. The clerk's twenty minutes alone at the queue are work; so are the team leader's twenty minutes behind the counter. A count of who stood at the window would see only one of them. The scene makes both visible by giving each a line. This reading of Daniels is this study's own; Daniels wrote of unpaid work at home and in institutions, not of a public counter.
The data-centre night scene, set in the 2010s and also synthetic, pushes the same point onto a pause that looks like idleness. A server has restarted itself and the operator sits down; a colleague sorts the alerts. Of its three readings, the scene leans to the one about who sorted:
The load differs depending on who sorted the alerts.
and its proposal closes with an instruction to make the filled time visible:
Bring out, too, who filled the time of the rest.
Hatton, in Mechanisms of Invisibility (2017), argues that work is made invisible through cultural, legal and spatial mechanisms that obscure the fact that work is done. The scene names a fourth, narrower kind that Hatton's list does not cover, and this is the study's own reading: work becomes invisible when one person's pause is booked and the colleague's covering is not. The scene does not say that the operator was resting rightly or wrongly. It says that a pause cannot be counted as empty hands before the covering is counted too.
Where the load gathers is as important to the scenes as that it moves. The script finds that nine of the thirteen speak, in their readings, judgment or proposal, of a load gathering on one person, on a few or on one side. In the QC laboratory scene, synthetic and set in a pharmaceutical plant of the 2020s, the rule is that whoever first sees an out-of-specification value writes the investigation report. The assistant manager who keeps finding them says so in numbers the scene supplies:
Of the six outlier reports this year, four are mine. My stability tests are three months behind.
Bolino and Turnley, in The Personal Costs of Citizenship Behavior (2005), found individual initiative associated with role overload and job stress; Bergeron, in The Potential Paradox of Organizational Citizenship Behavior (2007), proposed that time given to helping may come at the cost of one's own task performance, which reward systems favour. The scene stages exactly that trade, as a written supposition: the one who finds pays with the finder's own tests, while the analyst who puts ambiguous samples on again rather than reporting them avoids the month a report would cost.
The divers' shallow sea, a synthetic scene that spans the 1970s and the 2020s, carries the gathering of the load across fifty years. A cooperative once set the shallow water inside the rocks aside for older divers, so that doing less was a set share rather than a fault. Now most divers are old, and the scene's sanggun, the divers who go deepest, are three:
There are three of us to go into the deep. Our breath is this village's abalone price.
The judgment does not blame the older divers and does not praise the three. It says the arrangement kept its name and lost its balance:
The name of the assignment is the same, but its weight has turned upside down.
The scene is the clearest case in the set of a share that was once made fair by design and now concentrates a load without anyone holding back more than before. What has changed is how many people are on each side of the rule.
1.3A rate, a count, a rule
The second thing the scenes do is look for the cause in an arrangement before a person. Eleven of the thirteen judgments, by the script's test, set a person's motive or character aside in so many words: not laziness but a missing reference panel, not a shirker but the one who saw first, not a gesture of avoiding work but a seat without grounds. In four scenes the arrangement is a way of paying or counting, and in a fifth it is a rule.
The demolition site scene, synthetic and set in the 2000s, is the sharpest. A worker of fifteen years hears the wall change and steps back; the crew is paid by a contract rate per pyeong, a unit of floor area, and the hour stopped costs everyone. A second-year says what the rate makes of the act:
You're slipping off again, brother. At this rate the rate gets cut.
The foreman answers with a different name for the same act: the worker did not slip off but saw the danger first.
The scene's lean is its third reading, which does not ask whether the worker was right but who pays for the hour:
The problem is who pays for the hour stopped. Under a rate per pyeong, time stopped safely becomes everyone's loss, and only the person who said to stop gets the hard looks.
The judgment then follows the cost forward, to the next time:
If who pays for the value of the time stopped stays as it is, the next signal will not be spoken aloud.
This study reads the rate per pyeong, beside Ordóñez and colleagues' Goals Gone Wild (2009), as a goal with a narrow focus: what it counts is area torn down, and the hour that avoided a collapse falls outside it. The scene adds what the goal-setting critique, in its note, does not stress: the cost of the narrow focus is borne socially, as hard looks at the person who stopped. The scene is careful about its own premise. Its evidence calls the pressure of the rate an assumption and leaves open whether the worker's feel is always right; the scene is a written case, and it says so.
The helpdesk scene, synthetic and set in the 2000s, puts the count itself on stage. The desk is assessed by tickets handled, and the first words to every caller are to reboot and call again. The judgment refuses to call the advice the fault:
The reboot advice is not in itself passing work off. The count that tallies the same query coming back three times as three solved tickets makes the first words into passing work off.
This is close to the claim Smith, in On the Unintended Consequences of Publishing Performance Data in the Public Sector (1995), makes against indicator schemes: that they rest on inadequate models of production. This study reads the point this way: a model of production that counts a ticket as a unit of help cannot see that three tickets were one problem. The scene goes on to the cost of seeing it:
The senior's bundling of the queries from the same floor and going up was a judgment that looked at the cause rather than the number of tickets, and a judgment that actually loses out by the number of tickets.
The person the scene credits with looking at the cause is the one the count marks down. Where the load went, the scene says in its evidence, is out of the in-house talk and the ticket count: the time of three reboots stays with the caller's team.
The solar checklist scene, synthetic and set in the 2020s, turns the count into a form. Head office scores the crew by the share of checklist cells filled, and "not applicable" counts as filled. One reading says it outright:
The way of counting makes 'not applicable' the cheapest ending. The body that holds back came from the yardstick.
Bevan and Hood, in What's Measured Is What Matters (2006), name the assumption that a part can stand for the whole. The filled cell is such a part: it stands for a measured earth resistance, and the scene shows the gap between them opening into a stopped inverter two months later. Here the fit is close, but the scene keeps one thing the targets literature in its note does not have: an installer's line that a single tester per truck had gone to the next site. The scene's judgment rests on that missing tool as much as on the count, and its evidence leaves open whether bother was mixed in.
The courier dispatch scene, synthetic and spanning the 1990s and the 2020s, sets two ways of declining work side by side. On the radio, a rider could pretend not to hear; the dispatch manager knew, and squared the account with the next good job. In the app, a decline lowers the acceptance rate, and a low rate brings worse calls. The newest rider speaks for what follows:
I take all the rainy-day calls to the outskirts to fill my acceptance rate.
and the scene's lean names the mechanism:
When declining becomes an individual's number, only those who can afford to decline do so.
This scene sits oddly beside the work on social loafing, and the oddness is instructive. Karau and Williams, in Social Loafing (1993), count evaluation potential among the strongest moderators of reduced effort in groups. The app makes every rider's decline evaluable; on the loafing view as this study reads it, that should reduce holding back. The scene does not dispute that. It asks a different question: once declining is counted, who can still afford it, and where do the unwanted calls go? Its answer, as a written supposition, is the newest rider. The scene is also one of the two whose judgment does not set a motive aside, and it is even-handed about the past: it says the manager's accounts could tilt towards those close to the manager, so the old way cannot be called an even way.
The QC laboratory is the synthetic scene in which the arrangement is a rule rather than a count, and it is also where the scenes go furthest in saying that the holding back itself teaches. The young analyst who retests rather than reports says why:
When it's ambiguous I just put the sample on again first. If I write a report, a month goes up in smoke.
The judgment does not excuse the act and does not blame the analyst for it:
The third-year's way of running a retest first is something the rule taught. As long as the finder takes on 2–3 weeks, not finding becomes the rational side.
The short synthetic scenes make the same move with fewer words. In the colour room, the next shift's slowed hands are put down to a reference panel that went missing between shifts; at the auction inspection, the judgment asks who has the authority to fill a missing label before anyone is seen as avoiding it:
Before the inspection team leader is seen as a person trying to avoid it, who can fill in the place-of-origin label must be checked.
Across the four counts and the rule, the scenes stand nearer the second strand of the debate than the first. They read holding back as the visible end of an arrangement. What they do not do is deny that a person holds back. The next section turns to where they keep the person.
1.4Where the person comes back
The person who held back is not written out of the scenes; the scenes move the person to a later slot. In four of the thirteen, by the script's test, the counter-condition names a person's motive or standard as what would remain if the arrangement were mended and the holding back went on. In the repair counter:
If the counter kept calling the chief for every expensive estimate even after receiving the old part and the parts price list, the cause is not a lack of grounds but a wish to avoid anger.
In the solar checklist, the residue is named as bother; in the QC laboratory, as each person's own standard; in the synthetic night-study duty scene, where teachers buy and sell evening supervision, as goodwill:
If many teachers would stand the same number of times without taking a price, the price is not the difference in circumstances but only a way of recording goodwill.
The order matters. The scenes first ask what the arrangement did, and only if the arrangement were changed and nothing else changed would they let a motive stand. This is a test design, and it is close to what the loafing literature's moderators imply without saying: if evaluation potential and expectations of co-workers move effort, then a person's effort cannot be read before those are fixed. The scenes do not cite that literature and should not be taken to agree with it; the parallel is this study's reading.
Two scenes also say, against the drift of the rest, that holding back did harm. The QC laboratory's judgment ends by saying the price is paid outside the laboratory, and one of the solar checklist's readings names who paid:
Whoever filled the cell, the price of the equipment stopped two months later is paid by the customer and the team that goes again.
These are the two scenes set on the performance question where the written consequence reaches beyond the workplace. Their harm is placed, like the rest of the load, on a receiver: the patient the evidence of the laboratory scene mentions, the customer at the stopped installation. Even here the scenes do not make the harm a fault of character. They leave the decision of who answers for it to a record the scene cannot hold: the reporters and retests before and after, the checklists before and after more testers.
The proposals then answer the load where it lands. The script finds that eleven of the thirteen change a roster, a sheet, a count, a rule or an order, and eight divide, rotate or separate a share. In the demolition site:
Time stopped for a danger signal is taken out of the contract rate and paid separately by the prime contractor.
In the QC laboratory:
The person who finds an outlier writes only the first page of the report. The investigation is taken over by an investigation duty that rotates.
In the solar checklist, a separate "not measured" cell; in the courier dispatch, rainy-day outskirt calls counted apart from the acceptance rate; in the helpdesk, repeat queries bundled into one ticket. Each proposal moves the received load onto something that can be written down and shared. None proposes to watch the person more closely or to rate anyone. This study reads the proposals as part of the scenes' design, not as advice for any trade, and takes no view on whether they would help.
The two scenes whose judgments do not set a motive aside are both on the era question, and both put an automatic arrangement at the centre. The spraying drone, a synthetic scene supposed for 2030, asks who takes the paddy a drone skipped; the courier dispatch asks who takes the calls an app routes. In both, the scene's attention is so far on the arrangement that there is no person's motive to set aside, and the question is only where the skipped work goes.
Taken together, the scenes give holding back a fixed shape. A person does less; someone else does more and says so; the judgment names an arrangement; the counter-condition keeps a motive in reserve; the proposal moves the load onto a sheet. The shape is GAEKJU's, written for comparison. Its value for the debate is that it shows, case by case, what a reading that starts from the person would miss.