1.1Data and method
GAEKJU's public library held 496 questions when this study fetched it, each 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 asks one of three things of its kind of talk: whether it is the same across domains, whether it costs performance, or whether it has changed across eras. GAEKJU's English names for two of the kinds are "Rough talk" and "Their own code", and the library holds 59 and 58 synthetic scenes of them.
The choice of kinds follows from the earlier GAEKJU papers. A script read the case lists of all 8 of them, 101 cases in all, and found that they had read synthetic scenes of five kinds of talk and none of the two read here. The selection then follows one rule: every synthetic scene of the kinds of talk "Rough talk" and "Their own code" whose evidence holds no digit. The evidence is the part of a scene that says what is particular to 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. Applied to the whole library, it leaves 10 scenes, four of the first kind and six of the second, and sets 107 aside. The script stops if the case list differs from the rule's selection or shares a case with an earlier paper. Five scenes ask the performance question, three the era question and two the domain question.
Every scene is synthetic. Nine of the ten open their evidence with the label "Synthetic:". The tenth, the scene of the month-end push, is filed by GAEKJU as synthetic but differs in shape: its evidence opens without the label, it has three lines of talk where the others have two, and it is the only one whose readings run to two sentences. Six scenes are set in the 2020s, one in the 1980s, one in the 2010s, one moves from the 1980s to the 2020s, and one is supposed for 2030. The study reads each scene as a written frame, not as a report of what happens in a trade.
The scenes share one form: a setting, a short scene, lines of talk, a term, the evidence, three readings, a lean towards one reading, 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. Across the ten there are 21 lines of talk and 30 readings.
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 record; every block quote below is that translation, word for word. Roles are put neutrally, and no gendered pronoun is added where the Korean has none. GAEKJU's word for its scenes' proposals is kept in the source lines as "Asi".
The counts come from a script that reads only the fetched records and tests the Korean, with the words of each test listed beside the counts. It marks, in the rough scenes, the readings whose subject is the hard word and the reading that looks past it; in the code scenes, the readings that hold one code, word, sheet or cell to be enough and the readings that dismiss the short word; in both, the readings that call the words a signal, the reading each scene leans to, the judgments that say the words did not aim at a person, and what the proposals name. Each test was checked against the scenes by reading them, and the scenes each count rests on are listed with it. 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%.
1.2The hard word and what it points at
The rough scenes write a hard word and then look past it. The clearest case is the synthetic scene of the tangled order, set between reception and treatment in an emergency room. Waiting at reception and calls to the treatment room back up at once, and the head nurse speaks:
The order's tangled. It has to be put right again now.
The scene offers two readings about the sentence itself. One counts its cost, the other its use:
A hard word can raise the anxiety of patients and colleagues.
A hard word is a signal that makes the confusion known quickly.
It leans to neither. Another reading sets the hardness aside and names a gap in who may set the order, and the judgment follows that reading to its end:
The head nurse's words do not aim at a person; they point at the confusion of the order. Even so, what decides the harm in the situation is less the hardness of the words than the authority to reorder and the time to explain to patients.
The judgment makes two moves. It answers the question that Cortina, Magley, Williams and Langhout (2001) put to uncivil acts, the question of the target, and its answer is not a person but a confusion. Then it ranks the hardness below two other things, an authority and a time. The scene does not say the hard word was harmless; it says the harm is decided elsewhere, and it claims no outcome for any patient.
The pattern holds in three of the four rough scenes. Each offers two readings whose subject is the hard word, eight in all, and one that looks past it, four in all. Three of the four lean to the reading that looks past the word. Three of the four judgments say the words did not aim at a person or a colleague (in one, at a particular person). In the synthetic scene of the blank on the checklist, set in a substation's shift room, the shift leader's words are short and final:
This blank can't just be passed on.
The scene's readings divide as before: the bluntness brings checking forward, the bluntness makes the next doubt harder to voice, and the blank on the checklist itself makes the conversation. The scene leans to the blank. Its judgment again names a target, and again the target is a thing:
The target of the bluntness is not the colleague but the unchecked item. The disagreement is read as having come forward because the sheet and the field record could be seen together.
What the second sentence credits is not the bluntness but an arrangement: a sheet and a record that two people could look at at once. The colleague's answer offers that arrangement:
There's a separate field record from yesterday. I'll bring it and match it up together.
In all four rough scenes the second line of talk says what the speaker will do next: explain to those waiting, bring the record, check together, say how far one looked. Four of the six second lines in the code scenes do so as well, so the move is not the rough scenes' alone. Andersson and Pearson (1999) describe how incivility can spiral and examine the point at which a spiral starts. This study reads the rough scenes as writing that starting point and then giving the next line to someone who takes up the work rather than the tone. The reading is this study's own, and the scenes do not show what a later exchange would bring.
The synthetic scene of the night handover, set at a rolling mill in the 1980s, carries the same move to a stopped machine. The night foreman sighs roughly at a stopped machine, and the day shift's handover paper holds only the time of the stop. The foreman asks who, then, knows the rest. Of its two readings about the hard words, one calls them a reaction to the breakdown and the other a signal of a share of checking pushed onto the night. The scene leans to neither, but to the reading that the blanks in the paper handover must be filled before the meaning can be told apart. Its judgment names the target and then stops short of a cause:
In this scene the target of the rough words is not a particular person but an inspection range that cannot be known. With no record to tell the day's share from the night's, concluding the cause is put off.
Cramton (2001) and Weber and Camerer (2003) describe a turn from situation to person, in which a difficulty is put down to people rather than to the situation. The scene makes the opposite turn. It could have read the foreman's sigh as a judgment on the day crew; it places the target in an inspection range that cannot be known, because the paper records a time and nothing else. This study offers that contrast as its own reading of the scene against the two works.
The fourth rough scene is the exception: it is the only one whose lean falls on a reading about the word itself. In the synthetic scene of the raised voice, a pharmacy manager hears a complaint about waiting and raises their voice; the one dispensing has seen that a name on a bag differs and finds it hard to say so. This scene leans to a reading about the word itself:
After the voice is raised, the threshold for correction grows.
It is also the only scene in which the one who spoke loudly names it:
My voice was loud just now. Please look at the bag again first.
Yet the judgment, having granted the cost, places the decisive thing outside the voice:
The scene in which the correction stopped for a moment shows the possibility of a cost. Since the difference came out after the manager asked again, the channel through which disagreement can be voiced is decisive, more than people's moods.
The scene keeps both readings in play. It lets the raised voice carry a cost, and it lets a question asked a moment later carry what brought the difference out. It claims no dispensing error, and it says nothing about what would happen at any real counter. Of the four rough scenes, then, three look past the hard word, and the fourth looks at it and then past it.
One thing the rough scenes do not do. Two of their readings call the hard word a signal: in the tangled order, a signal that makes the confusion known; in the night handover, a signal of a share pushed onto the night. The scenes lean to neither. The words are heard as pointing at something, but their hardness is not counted as the information.
1.3The short word and the account beside it
The code scenes set a short word beside the longer account it leaves out, and they decline to let either stand alone. Five of the six offer a reading that one code, one word, one sheet or one cell is enough, a reading that takes the short form to need nothing beside it. None leans to it. The synthetic scene of the grade code, set in the holding pen of a cattle auction, opens with the code at its most efficient: the auction officer says that with the code the price sheet comes out at once. The farm support officer answers with what the code does not carry:
I'll attach the state of the legs beside the sheet as well. It's a part the one who watched for a long time knows.
The scene's readings run from the view that the experience of care helps the trade from outside the grade table to the view that the same code means the same information is needed. It leans to the first. The judgment keeps both kinds of talk:
Grade talk is needed for the trade, but it cannot hold all the information from the time the animal was cared for. If the two kinds of talk are set side by side, the difference can be seen without marking one side down as unprofessional.
Here the code belongs to the formal side, and the longer account to the one who raised the animal. Bowker and Star (1999) hold that each standard and category valorises some point of view and silences another. This study reads the judgment as an answer to that claim in miniature: the grade keeps the point of view of the trade, the note on the legs keeps the point of view of care, and the scene asks for both at one place of judgment rather than for a better grade. The last clause, "without marking one side down as unprofessional", names the cost the scene wants to avoid; this study reads it as a cost that would fall on the longer account rather than on the code.
In other scenes the short word belongs to the floor, and the formal form is the one that leaves things out. The synthetic scene of the new nickname moves from the 1980s to the 2020s at a machine-tool factory. The meanings of a crew's nicknames for frequent stoppages are put onto the official work sheet, and the crew then gives a stoppage the sheet does not hold a new nickname. The scene offers the reading that the official sheet makes nicknames useless, and the reading that nicknames are only habit, and leans to neither:
A word once written on the sheet is not the end. If the new nickname was put on a stoppage different from before, the colleagues saw first a condition that has no name yet.
Orr (1996) found that the work of technicians holds elements not contained in the official definition of the job yet vital to its success, and that their knowledge travels mainly in talk. This study reads the scene as giving that finding a motion over time. The official sheet absorbs the earlier nicknames, and the talk moves on to what the sheet does not yet hold. The scene claims nothing about any factory's change of language; it says so in its evidence.
The synthetic scene of the ambiguous answer, supposed for 2030, puts the short word against the one item that an automatic judgment picks from a set classification. The lead of the evaluators speaks of a model's answer:
It's ambiguous. Rather than wrong, it sits across a boundary.
The automatic judgment picks one set item and puts it forward. The scene offers three readings: one item should be chosen at once, "ambiguous" is said for not knowing the standard, and "ambiguous" is a signal of having seen a boundary case first. It leans to the third:
The evaluator's hesitation may be not a mark of ignorance but information that the boundary of the classification is narrow. Until what difference was seen is compared with other evaluators, it is not closed into one item.
The word "ambiguous" is the shortest form of a judgment that has not settled. A cell in a table is shorter still. The scene sets them side by side and keeps the table open until the word can be explained. Star and Griesemer (1989) count standardised forms among the boundary objects that keep their identity across viewpoints; this study reads the classification table here as a standardised form that the scene does not trust to hold the evaluator's viewpoint without a sentence beside it. That reading goes beyond the work.
The synthetic scene of hold and wait has two short words for one day. A supervision manager says the check of the drawings is on hold; the builders keep their workers on site and call it waiting. The manager explains the hold as scope left to check, and the construction lead answers:
Meanwhile our workers are waiting. Please give the time for moving them to another work stage.
One reading says the two words mean the same stop. The scene leans instead to the reading that a hold is the supervision's judgment and a wait the builders' cost, and the judgment sets matching the words aside:
Even on the same day, it is checking time for the supervision and tied-up workers for the builders. Rather than matching the words, one has to bring out at what point who can reassign the workers.
Bechky (2003) located misunderstanding between occupational groups in differences of language, of the place of practice and of how the product is conceived, and found that solutions invoking those differences could transform understanding. This study reads the scene as keeping the two words apart and asking for the difference in the place of practice to be brought out: who holds the workers, and when they can move.
In the synthetic scene of the abbreviation the short word is a patent attorney's, and one reading holds that it saves consultation time and catches the core. The scene leans to the reading that which way it is cannot be known until the drawing comes out, and the inventor answers by proposing a drawing:
A drawing will be faster than words. It bends once more here.
Carlile (2002) describes a boundary object as a means of representing, learning about and transforming knowledge at a boundary. This study reads the drawing as working in that way between a legal scope and a moving part; the judgment waits for it: only after seeing the drawing can the value and the loss of the compression be judged.
The synthetic scene of the month-end push, the one that differs in shape, gives the short word the most weight. "Pushing out" is the floor's word for the rush to meet another firm's monthly closing, when outgoing inspection runs on reduced samples. A quality manager has marked the dates of returned parts on a calendar:
It has to be shown that it's not the people but the calendar that is shaped that way.
The scene leans to the reading that the word names the pressure exactly:
'Pushing out' is a word that names exactly the shape of the ordering firm's closing pressing on the partner firm's inspection time. The month-end returns come from the calendar more than from people.
It does not leave the other readings unanswered. One holds that the partner firm accepted this way of trading and did not add inspectors; another that the ordering firm's own incoming inspection grows tighter at the same time. The judgment names a way to tell the last apart, laying the count of incoming inspections over the same calendar, and leaves it undone. The scene's sentence about people and the calendar is the plainest statement in the set of the situational reading that, in Cramton's (2001) account and in Weber and Camerer's (2003), is put aside in favour of the person. This study notes that the calendar is a loose fit for a boundary object in Star and Griesemer's sense: it is carried from one firm to another to change a view, not to hold two views at once.
Across the six code scenes, then, the leans fall on readings in which no one short form holds everything. Two of those leaned readings call the short word a signal: the new nickname as a signal of a stoppage not yet classified, "ambiguous" as a signal of a boundary case seen first. Set beside the rough scenes, where two signal readings are offered and none is leaned to, the difference is the clearest finding of this section. Of the four readings across the ten scenes that call the words a signal, the scenes lean to two, and both concern short words.
1.4What the proposals change
No proposal of the ten asks anyone to change the words. None asks for a softer tone, a lower voice, a code dropped or a term made standard. What the proposals change is time, place and authority. Nine of the ten name a time, four give an authority or an option, and seven set two things together or side by side, five of the six in the code scenes and two of the four in the rough scenes.
The tangled order, the scene whose hard word is the plainest, answers it with an authority and a share of time:
Set clearly, between reception and treatment, the authority to set the order again. Share out the time to explain to patients together with the work of calling.
The ambiguous answer, the scene whose short word is the most hesitant, answers it with a review time and a return of authority:
Move boundary cases to a separate review time. The automatic judgment shows side by side the reason it picked one cell and the difference the person left, and returns the authority to revise to the evaluators.
Two proposals, both in rough scenes, refuse to pile checking or explaining onto one person. The blank on the checklist divides the authority to pass on, so that the responsibility for checking does not fall on the one colleague who found the answer. The raised voice, the one proposal that names no time, separates explaining the wait from checking the dispensing:
Do not lay explaining the wait and checking the dispensing over the same person. Leave room in the order so that the one who saw the difference on the bag can ask for a check at once.
This is the scene that leaned to the cost of the raised voice, and its proposal does not touch the voice. This study reads it as moving the pressure the voice passed on away from the one who has to check. The scenes, in short, treat the surface of talk as a pointer and make their changes where the pointer leads.