Research cases on what workplace talk reveals about an organization, built with AI and checked across model families. Cases from other workplaces and eras cite published research. No particular company or person is identified or portrayed.
Talk that never went up the chain. Choices to rebalance.
Rough talk, initialisms, workplace jargon, talk about the bosses, slacking, passing the buck, wasted effort, mistakes, pushback and resignation, and the small talk that fills every day: what everyone below the chief executive, team leads and middle managers included, said to one another. GAEKJU traces structural imbalances in that talk and asks whether they really cost performance. The questions bring the imbalance to light and put each choice up for re-examination.
Human slack, waste and error do not necessarily lead to low performance.
“Necessarily” falls to one counterexample, so the proof is a collection of counterexamples, from the cases in this library and from published studies of other workplaces and eras, with the evidence against printed beside them. What remains is the real question: when does it hurt?
- 01 · Other workplaces, other eras1959–2014
A machine crew that got through the day on horseplay (1959). Wards with better relationships reported more drug errors (1996). A packing team whose rough talk carried solidarity (2004). A call centre that got faster when teams took breaks together (2010). Output that rose behind a curtain (2012).
- 02 · Evidence against1913–2013
Effort that shrinks as more hands share the rope (1913, 1979, 1993). Incivility aimed at people, which cut the effort and work quality of those on the receiving end (2013).
What remains: when does it hurt? Performance cards
The questions we asked about each defect, and the workplaces we carried them into.
Drawing the map…
—
Questions on this map. Squares are record readings, diamonds are outside cases, circles are synthetic scenes in other trades. A dashed line marks a pattern moved to another trade; a solid line marks a comparison with an outside case.
Distance on this map does not indicate similarity or causation.
The same behaviour, a new name every generation.
In 1911 the name came from a manager who wanted it gone. In 1952 and 1959 it came from a researcher who ran the machines beside the crew. The behaviour barely changed; the names, and who gave them, did.
“Soldiering”: working slowly on purpose, to be stamped out by time study
Taylor, an engineer on management’s side ↗A wiring crew sets its own fair day’s output and sanctions the rate-buster, the chiseler and the squealer
Hawthorne studies“Goldbricking”: slowing down on badly priced piecework, stopping at the quota on good work
Roy, working the machines himself ↗“Banana time”: a crew breaks the day into teasing rituals, and cannot bear the day when the play stops
Roy ↗Exit, voice, loyalty
Hirschman“Social loafing”: each pair of hands does less as the group grows
Latané, Williams & Harkins ↗“Neglect” joins the list: lateness, absence, more errors
Farrell ↗“Weapons of the weak”: foot-dragging, feigned compliance, gossip
ScottBehind a curtain, out of the managers’ sight, output rises 10–15%
Bernstein ↗“Quiet quitting”: doing only what is asked
Gallup survey ↗
Question the structure, never the person.
- Research cases on what workplace talk reveals about an organization, built with AI and checked across model families. Cases from other workplaces and eras cite published research. No particular company or person is identified or portrayed.
- Whether a reading holds in another workplace is a hypothesis to test.
- Where a case on this page has a link, it leads to the published study itself.
- Each card carries the author’s judgment, its as-of date and the condition that would overturn it. Judgments can be revised.
- A defect in the talk is read as a sign of the structure around it, not as a trait of the person who spoke.
One question, three kinds of evidence.
Start with a record reading, compare it with outside research, then test it in an imagined setting. Each kind of evidence has its own limits.
Open a record reading
A record reading shows where, and how often, a kind of expression appeared. It does not show what anyone said in full or why they said it.
Browse cards 02 / Compare an outside caseCompare an outside case
Published studies test whether a similar pattern appears in another workplace or era. Their findings are kept apart from the cards here.
Browse cards 03 / Try a synthetic sceneTry a synthetic scene
An authored scene moves the question into a new setting. Its dialogue is fictional and its outcome is a hypothesis to challenge.
Browse cardsExplore further
Question library
Questions on seven kinds of talk: record readings, outside cases and synthetic scenes.
OpenQuestion atlas
Every card as a dot, grouped by kind and question.
OpenScale mode
The same cards in WebGL: fly, search, zoom.
OpenSocial matrix
525 synthetic hypotheses across 25 social domains.
Open