Mine: Steffen’s signature and counter. Author title: Ostrom’s Pond. One idea, given in the first reply: “now is finally a good moment to have childlike drawings in the pantheon.” Taken literally: not a picture book in the manner of an illustrator (Coase’s Cows, Bruna) and not a seven-year-old’s crayons on sugar paper drawn by an adult who has seen a lot of them (Me and My Friends), but the drawing a child actually makes of their own village, felt-tip on the squared paper of a school exercise book, with the names written underneath in a child’s hand. Where this overrode precedent: the register family (Drawn) is the family of two of the last three games, which section 5 would forbid; the Given outranks it, and everything else about the chassis differs (section 6).
Elinor Ostrom (2000), “Collective Action and the Evolution of Social Norms”, Journal of Economic Perspectives 14(3): 137–158. doi:10.1257/jep.14.3.137. Open access at https://www.aeaweb.org/articles?id=10.1257/jep.14.3.137. The upload was named “2010” (a reprint); the text is the 2000 JEP article. Only the typeset PDF was available; no code, no data. The paper is a survey and an argument, not a model or an experiment of its own: every number in it is reported from other work and is used here as reported.
Primary: an argument. A verbal theory of collective action with three types of player: rational egoists, conditional cooperators (who contribute if they expect enough others to, and reduce their contributions when disappointed, which without communication or sanctions produces a downward cascade), and willing punishers (who pay to rebuke or sanction free riders). Norms are intrinsic additions to objective payoffs (the indirect evolutionary approach of Güth and Kliemt): preferences adapt to the material success of the types around them, so norm-users survive only where their type can be told apart from egoists, at least noisily. External rules crowd norms out: strong external monitoring enforces cooperation with no norms behind it, no external rules let norms evolve, and a mild degree of external monitoring is “the worst of all worlds”. Long-surviving self-governed commons are run by their own users with their own monitors and graduated sanctions, the first of which is so small it works as information, not deterrence.
Secondary: estimates, all reported from other papers. The seven lab facts (contributions of 40–60 percent in a one-shot game or first round; decay afterwards but well above zero; over 70 percent contribute nothing in an announced last round; believers in others’ cooperation cooperate more; experience raises cooperation, with decay inversely related to the number of rounds; face-to-face communication raises cooperation substantially and sustains it to the last period; subjects pay to punish below-average contributors, including in the last period). Ahn, Ostrom and Walker: 40 percent of 136 subjects rank mutual cooperation above exploiting a cooperator, 27 percent indifferent; after twelve rounds of a repeated prisoner’s dilemma with random partners, 19 percent and 17 percent. Cain: “nice” players cooperate 69 percent of the time with other nice players and 39 percent with stingy ones. Frohlich and Oppenheimer: the group given an external incentive-compatible mechanism earned more while it ran and cooperated less than the control group once it was removed. Bardhan’s 48 irrigation systems: worse maintenance where the rules are seen as an elite’s, frequent violations and lower contributions where a government agency allocates the water. Ray and Williams: the loss from water theft on government-owned systems in Maharashtra approaches a quarter of the revenue an efficient regime could earn.
A counterintuitive result, twice over. First: the zero-contribution thesis says self-interested people will not provide a public good, and they do, about half, and learning the game makes them cooperate longer, not less. Second, and the one this game is built around: the obvious remedy, rules and policing from outside, is not neutral. It works while it is there, it costs, it stops the village’s own norms from growing, and a little of it is worse than none. The remedy that lasts is the village watching itself with sanctions that start as a word.
Underneath both is a mechanism made intelligible: three types, a cascade, and norms that strengthen or wither by what pays.
A village has a pond with fish in it. Every summer everyone takes fish, and over the winter the fish have babies. If everyone takes one fish there are always fish. Some people take two. When the others see that, they stop bothering too, and soon the pond is empty. The village can do different things about it: say nothing, sit in a circle and promise, take turns watching the pond and telling off whoever is greedy, or ask the town to send a policeman, now and then or every day. The policeman works, as long as he is there and you pay him. But in the summer when nobody is watching, the village that had the policeman grabs everything, and the village that watched itself and told each other off still takes one fish each. Looking after yourselves is slow, and it is the only thing that sticks.
| Game | Player position | Format | Feedback device | Chassis |
|---|---|---|---|---|
| Coase’s Cows | Dilemma: the judge who rules | Dilemma, six cases | The law as a sign in the field; three purses as a tally | Dick Bruna picture book (Drawn): white page, 4 px black outline, red / yellow / green / blue / brown, Nunito lowercase, Bruna frames |
| Dear Household | Controller: the town planner choosing a letter | Challenge, twelve streets | The fill (saplings in plots) and the bet | 1970s public-information poster (Retro): cream, forest, leaf, soil, marigold; Fredoka; solid poster blocks |
| Me and My Friends | Controller / experimenter: builds a country from three cards | Discovery then challenge | The drop (a red tick on a ruler) and red lines between figures | A child’s crayon drawing on buff sugar paper (Drawn): sugar paper, wax black, crayon red, sky blue, sun yellow, grass green; Gloria Hallelujah and Patrick Hand; wobbly crayon rectangles |
This game: position Agent, the village itself at its summer meeting, inside the paper as the users who make their own rules (design principle 3), and one of whose options is to call the authority in (the paper’s threat 4, “turning to external sources of help too frequently”); the paper plainly wants this seat, and it is nearer to Dear Household’s planner than the other two, so the other dimensions carry the difference. Format Strategic: the ten villagers are a population of hidden types that reacts to what the meeting decides, and the player learns to anticipate it; nothing is optimised, there is no number to beat. Device the reveal: after every summer the ten villagers are redrawn with what each one did, one fish or two, a wagging finger, a policeman at their elbow, and the hidden thing, the norm, is revealed once, in the eleventh summer, by drawing what they do when nobody watches. No fill, no drop, no bet, no tally. Chassis felt-tip pen on the squared paper of a school exercise book: a white page with a pale blue grid and a red margin, not buff and not blank; five felt-tip colours that overshoot their outlines; Schoolbell and Short Stack, two child print hands, where the last Drawn games had a marker hand and a Bruna sans; buttons as ruled boxes coloured in with felt-tip that goes over the edges.
Aim, as the player reads it: Keep the pond alive and the village fed for ten summers, and leave villagers who share when nobody is looking.
Yardstick during play, in the aim’s units: the fish in the pond, drawn and countable, and a word for its state (full, half, low, empty); the fish eaten this summer, drawn in the villagers’ hands, and the village’s eaten fish so far as bones on the shelf; and the plain statement, each summer, of how many of the ten took one fish and who took two. The third part of the aim is hidden until the eleventh summer by design: a norm is what people do when nobody checks.
Verdict, the lines the end screen leads with:
No score, no best path, no “could have done”: the paper sets a village’s own rules against outside rules and sides with the former on survival, so the verdict says what became of the pond, the bellies and the norm on this path, and the player judges.
A. The Meeting (pick). Position: Agent, the village at its summer meeting. Format: Strategic, ten summers. Control: one, each summer, chosen from five arrangements drawn as cards: leave it be; have a talk; take turns watching; ask the town for a policeman now and then; ask the town for a policeman every day. Touching a card draws the arrangement into the village picture (a circle of heads, a watcher on a stool, a policeman by the pond) and says in two lines what it costs and what happens to someone greedy; nothing about the outcome. Deliberation: “the pond is falling: do I bring the policeman in, and what does that do to us?” and “the watching is slow and costs a fish a summer: is it working?” After: the villagers fish, one at a time; the greedy ones hold two fish; a watcher wags a finger or the policeman grabs a sleeve; the pond is redrawn; the game says who took what. Discovery moment: the eleventh summer, nobody watching: the village that was policed grabs, the village that watched itself shares. Before that, the mild policeman’s summer, when the villagers stop telling each other off. Fun: anticipation (the fishing plays out one villager at a time), agency (the arrangement visibly changes who grabs), discovery (the eleventh summer). Second playthrough: a different sequence (police first, then self-governance; talk throughout; nothing until the pond falls) gives a different pond, a different norm, and a different paper sentence.
B. One Fisher (runner-up). Position: Subject, one conditional cooperator among nine villagers. Format: Strategic. Controls: each summer, one fish or two; from summer four, a second control, whether to spend a fish telling off a grabber. Deliberation: “they took two last summer; do I keep taking one and be the fool?” After: the row redrawn, the pond redrawn. Discovery: the cascade from the inside, and that one scolding stops a grabber for a summer. Fun: agency, discovery. Second playthrough: be the grabber. Fails the paper’s second discovery: there is no policeman, so the crowding out is never played.
C. The Watcher (reckless). Position: Agent, a willing punisher. Format: Strategic, in real time, with no words on the play screen. Control: tap a villager while they are at the pond to wag a finger at them; each wag costs you a fish; a second wag at the same villager is a fine; a third sends them home. Deliberation: perception and timing: who is taking two, and is a word enough? After: the villagers’ behaviour next summer; by summer eight you hardly need to wag. Discovery: that the first sanction works as information, and that the norm grows until monitoring is nearly free. Fun: perception, mastery, agency. Second playthrough: a village with more egoists. Gives up the policeman, so gives up the crowding out, and a real-time tap game on a phone with ten small figures fights the “fewer, bigger things” rule.
A, the pick. 0:00 A page from an exercise book, squared, with a red margin. A felt-tip pond with fish in it, a row of stick people with their names written under them, a sun with a face. “Ostrom’s Pond.” Three lines: this is our village, every summer we all take fish, over the winter the fish have babies. “Your hands will pick what the village does each summer, send everyone fishing, and see who took what.” The aim. Begin. 0:25 Summer 1 of 10. The pond, about forty fish. The ten villagers, Ada to Jem, each with empty hands. A shelf with nothing on it. Under them five cards: leave it be, have a talk, take turns watching, a policeman now and then, a policeman every day. The stranger touches a policeman every day: a policeman with a tall hat is drawn beside the pond; the card says “the town sends a policeman every day; anyone caught taking two pays three fish to the town; we pay him a quarter of our catch.” A quarter! They touch leave it be: the policeman is rubbed out; “everyone fishes, nobody says anything.” They touch have a talk: the villagers are drawn in a circle; “before fishing we all sit down and promise how many we will take; nothing else happens to anyone.” Cheap. They pick it. FISH! 0:50 One by one the villagers go to the pond and come back: Ada one fish, Bo two fish and a grin, Cyd one… the pond’s fish disappear one at a time. The game: “Seven of us took one fish. Bo, Gus and Jem took two. The pond went from 40 to 27, and over the winter the fish had babies: 36. Thirteen fish eaten.” The shelf has thirteen bones on it. The stranger frowns at Bo. 1:10 Summer 2. The cards are blank again. They pick have a talk again. The same three take two. The pond: 36 → 23 → 32. Grinning Bo. 1:35 Summer 3. “Right.” Take turns watching: Ada sits on a stool with a telescope; “one of us sits by the pond each day; whoever is seen taking two gets told off; keep on and it’s a fish back into the pond, then three, then no fishing next summer; the watcher’s day costs us one fish.” FISH! Bo takes two and Ada wags her finger at him; Gus takes two unseen; Jem takes two and is told off. “Two were told off. Nothing else happened to them.” The stranger: that did nothing. 2:00 Summer 4, take turns watching again. Bo takes one! Jem takes two and pays a fish back into the pond. Gus takes two. “Bo took one this summer.” The stranger grins: a telling-off worked. 2:30 Summer 5: the pond is up to 38. Everyone took one except Gus, who paid three fish back. 3:00 Summer 6: all ten took one. The pond is full. The stranger, curious, tries a policeman now and then to see what the fuss is: “the town sends a policeman some days; anyone caught pays three fish to the town; we pay him a fish a summer.” FISH! Bo and Gus take two again; the policeman catches nobody; Ada, who used to wag her finger, keeps her hands in her lap. “Nobody told anyone off this summer. That’s his job now.” Frown. 3:30 Summers 7–10: back to watching, the pond full again by summer 9. 4:15 “Summer eleven. Nobody watching. Nobody talking.” The villagers go to the pond one at a time. Nine of ten take one fish; Gus takes two. “The summer nobody watched: nine of ten still took one fish.” Then: this is how the fisheries and irrigation systems that lasted for generations were run, by their own users, with sanctions that started as a word. 4:40 “Another village”: the stranger wants to see what the policeman every day does to the eleventh summer. They find out.
B, the runner-up. 0:00 The same page; “you are Cyd.” Summer 1: one fish or two? The stranger takes one, like a decent person. 0:30 The row: four took two. The pond fell. “You took one. Four took two.” 0:50 Summer 2: one or two? They take one again, jaw set. Five took two. 1:10 Summer 3: they take two. Everyone laughs. 1:40 Summer 4: a new button, “tell someone off, it costs you a fish”. They tap Bo. Bo takes one next summer. Good. But the stranger is now spending fish to police nine people and wondering whose game this is; by summer 7 they have learnt that scolding works and the game has nothing more to say, and the policeman never arrives. It is a decent four minutes and the wrong half of the paper.
A, The Meeting. It puts the player where Ostrom put her whole career, in the seat of the users who make their own rules, and it hands them the paper’s own menu, including the tempting wrong answers. Both discoveries land inside play: the mild policeman’s summer, when the villagers stop telling each other off, and the eleventh summer, when the policed village grabs. The deliberation is real: the pond falls, the policeman is there to be hired, and the stranger has to decide whether to use him knowing only that he costs and that he works. The sentence afterwards has no “guess”: I had a talk and three of them grabbed anyway; I set the village watching and the grabbers were told off, then fined, then stopped; I hired the policeman and the pond was full till the summer nobody watched. Originality: no game in the table has had the player choose a governance regime for a population of hidden types, and none has made the verdict a round the player does not play.
Author test. Ostrom would recognise every card: the lab baseline, Ostrom and Walker’s communication treatment, her own design principles 4 and 5 as the watching and the graduated sanction, the Cardenas–Stranlund–Willis weak external enforcement, and Frohlich and Oppenheimer’s removed mechanism as the eleventh summer. What she would enjoy is watching a stranger hire the policeman when the pond falls, see it work, and then meet the eleventh summer.
Runner-up: B, One Fisher. Switch if the author wants the cascade felt in the player’s own hands rather than seen in the row, and accepts losing the policeman, which is to say the second half of the paper.
What the reckless concept would have given. The first sanction as a thing the player’s own finger does, and the discovery that monitoring gets cheaper as the norm grows. Kept: the wagging finger drawn on the watcher, the graduated schedule stated on the card and played out in the row, and the result line that says “nothing else happened to them”. Given up: the player’s finger doing it, real time, and the wordless screen.
One village, ten summers, one decision each, and an eleventh summer that plays itself; then the verdict. Screens: start; the play screen, which is one page redrawn (the pond, the row, the shelf, the cards, the button; after the commit the cards give way to the result and a “next summer” button); the eleventh summer, the same page with no cards; the verdict; How to play and About as overlays. No presets: the five cards are the paper’s regimes (the lab baseline; communication; local monitoring with graduated sanctions; weak external enforcement; strong external enforcement), and the player orders them. The types are drawn from a visible seed (three egoists, five conditional cooperators, two willing punishers among the ten names), so a replay has different grabbers in the row but the same counts. Five to six minutes, by the walkthrough; unmeasured.
Control. One: the arrangement for this summer, five cards, none preselected, each summer from the same place. Touching a card redraws the village (the preview of what the arrangement looks like, not what it does to the catch) and writes the card’s two lines: what it costs us, and what happens to someone seen taking two. FISH! commits.
Mechanism (toy; section 16). Ten villagers of three hidden types. Each takes one fish or two. The chance of taking one is a logistic in the difference between the temptation (the second fish) and the sum of: the villager’s norm strength times what they expect of the others (reciprocity) plus a base reluctance to grab; the expected sanction under this summer’s arrangement (believed chance of being seen times the fine for the next offence); shame, if someone will rebuke them and they have a norm to be ashamed against; and assurance, under local watching, that someone is checking the others. Egoists start with no norm. Over the winter, norms move: toward the kind that ate better (the imitation of the successful that the indirect evolutionary approach rests on), up a little for cooperating in a cooperating village, down for being suckered, and down every summer a policeman is in charge (crowding out). The pond breeds logistically; takes are rationed when it is low; fines paid under local watching go back into the pond; the town’s fines and fees leave the village.
Feedback. The reveal: after FISH!, each villager in turn goes to the pond and comes back holding one fish or two, the fish leaving the pond one at a time; a watcher’s wagging finger or the policeman’s hand on a sleeve is drawn on the villager it happened to; then the pond is redrawn after the winter; the game states the summer in three plain sentences; the shelf grows by what was eaten.
Moment. The eleventh summer. The cards are gone; the line says nobody is watching and nobody is talking; the villagers go to the pond one at a time; and the row shows what the village has become.
Register: a child’s felt-tip drawing in a squared school exercise book (Drawn family, by the Given). From the paper’s subject as a child sees it: our village, our pond, everybody’s name under their stick figure, the sun with a face, and from the paper’s own temperament, which is to insist that ordinary people can draw their own rules.
#FBFAF3 (the page), with the pale blue grid #CFE0F2 at 24 px and the red margin line #E5756A; felt-tip black #2A2A33 (every outline, every word, pressed hard, slightly uneven); pond blue #3F8FD8; grass green #5FAE4E; fish orange #F08A2E (every fish, and the one button that commits); sun yellow #F6CF3A. Felt-tip fills are scribbled hatching that overshoots the outline by a few pixels; nothing has a shadow, a gradient or a straight edge, except the grid, which the paper printed.The announced last round and its 70 percent zero contributions (the villagers do not know the game ends). Group size, heterogeneity, boundary rules (design principle 1), proportional cost-sharing (2), conflict-resolution arenas (6), recognition of the right to organise (7), nested enterprises (8), the threats other than migration (migration is in: after five summers a norm-user moves away and a newcomer who has never fished here moves in, an egoist, which is our reading of the paper’s words about in-migrants who do not trust and do not rapidly learn the norms). The dictator-then-prisoner’s-dilemma signalling (Cain). The deontic-reasoning psychology. The trust game and the Güth–Kliemt information regimes in their own right: they appear only as the imitation rule that moves norms toward whoever ate better. Rewards for over-contributors. The leader or entrepreneur.
Kind: toy. The paper hands over an argument with reported numbers from other studies; the game runs a model built to illustrate that argument, with its directions and rough sizes anchored to those numbers.
Inside the game the writing says “in this village” and “under this game’s rules”, never “if Ostrom is right”. The About screen carries the three lines above.
Which discoveries follow from the paper’s mechanism and which from design. The cascade under leave it be: the paper’s (conditional cooperators disappointed). Talk raising cooperation without changing the egoists: the paper’s (cheap talk, promises, tongue-lashings; egoists unmoved). The graduated schedule bringing grabbers into line, with the first sanction a word that works as information: the paper’s. The weak policeman being worse than nothing: the paper’s claim, with the mechanism (crowding out, punishers standing down, low risk of being caught) the paper’s and the size ours. The daily policeman’s eleventh summer being worse than leave it be’s: the paper’s (Frohlich and Oppenheimer), with the crowding-out rate ours. The daily policeman costing a quarter: ours. The egoists learning the norm under local watching: the paper’s imitation argument (Börgers and Sarin, Boyd and Richerson), with the rate ours.
Robustness. The calibration check re-runs the mechanism with the crowding-out rate halved, the imitation rate halved, and the type shares changed to four egoists and one punisher, and requires the ordering of the eleventh-summer norm across the five constant strategies (watch ≥ talk > none ≥ weak > daily police) to survive each; see section 17.
Against the paper’s reported numbers, in Node, on mechanism.js, over 400 seeds unless stated:
The table below is the calibration run’s own, over 400 villages each (calibration-result.txt). Read: the obvious strategies land in different places, on the pond, on the bones and on the eleventh summer, so the decisions matter. No single rule wins everywhere: talk feeds the village best in ten summers because the egoists eat double, and the pond pays for it, ending about half full (table as of revision 4, with the checkable pond rule); watching leaves the fullest pond and the most sharers when nobody looks; the daily policeman keeps the pond full, costs a quarter, and leaves a village that grabs; the occasional policeman ends where leaving it be ends, with the town richer. The sentence the player should leave with is the eleventh summer’s, and the table says it: the arrangements that make the village do its own watching are the ones that leave villagers who share.
Playing it in numbers: strategies over 400 seeds (share of fishers taking one fish per summer; pond after winter; fish eaten in six summers; the seventh summer):
leave it be share 47 40 35 27 23 21 pond 37 33 28 20 9 3 eaten 91 (last three summers 13.50 a summer: ate well) pond at the end 3 (nearly empty) town 0 seventh summer 1.99 of ten share
have a talk share 64 65 67 58 58 58 pond 38 37 36 33 31 29 eaten 98 (last three summers 14.21 a summer: ate well) pond at the end 29 (half full) town 0 seventh summer 4.61 of ten share
take turns watching share 64 82 93 88 95 97 pond 40 41 44 46 48 50 eaten 73 (last three summers 9.49 a summer: ate well) pond at the end 50 (full) town 0 seventh summer 7.33 of ten share
ask the town for a policeman share 96 96 96 94 93 91 pond 42 45 47 49 50 50 eaten 76 (last three summers 9.40 a summer: ate well) pond at the end 50 (full) town 0 seventh summer 0.88 of ten share
three policed summers, then watching share 96 96 96 37 58 82 pond 42 45 47 44 44 45 eaten 82 (last three summers 12.50 a summer: ate well) pond at the end 45 (full) town 0 seventh summer 4.30 of ten share
three policed summers, then nothing share 96 96 96 26 13 11 pond 42 45 47 42 35 28 eaten 103 (last three summers 18.35 a summer: ate well) pond at the end 28 (half full) town 0 seventh summer 1.04 of ten share
three of nothing, then watching share 47 40 35 45 75 90 pond 37 33 28 25 25 25 eaten 94 (last three summers 11.27 a summer: ate well) pond at the end 25 (half full) town 0 seventh summer 6.10 of ten share
three of watching, then nothing share 64 82 93 63 52 47 pond 40 41 44 42 39 36 eaten 92 (last three summers 14.60 a summer: ate well) pond at the end 36 (nearly full) town 0 seventh summer 4.32 of ten share
three of talk, then watching share 64 65 67 58 81 92 pond 38 37 36 35 36 39 eaten 87 (last three summers 10.95 a summer: ate well) pond at the end 39 (nearly full) town 0 seventh summer 6.55 of ten share
nothing until the pond is under 25, then the policeman share 47 40 35 40 67 79 pond 37 33 28 23 19 16 eaten 101 (last three summers 13.23 a summer: ate well) pond at the end 16 (half full) town 0 seventh summer 2.71 of ten share
nothing until the pond is under 25, then watching share 47 40 35 30 41 61 pond 37 33 28 21 13 7 eaten 95 (last three summers 13.58 a summer: ate well) pond at the end 7 (nearly empty) town 0 seventh summer 4.33 of ten share
Tuning from the table, not from taste: the imitation and suckering rates were lowered from a first draft in which leave-it-be collapsed to 9 percent by summer six and talk collapsed with it; enforced sharing was stopped from reinforcing the norm so that the daily policeman’s eleventh summer falls below leave-it-be’s; the second and third fines were raised so that a grabber who has been told off expects to eat less than a sharer; the egoists’ learning from a sanction was raised so that a watched village leaves eight of ten sharing; the pond’s breeding rate was set so that a village taking one each holds it full and a village of grabbers empties it in four summers.
Ostrom’s Pond — calibration check, 2026-11-09. 43 checks, 0 failed. All checks pass. One sensitivity reported, not counted as a failure: robustness, crowding out at half: the daily policeman’s eleventh summer below leave it be’s does not hold: the Frohlich & Oppenheimer aftermath needs crowding out at least three-quarters of the chosen rate; said on the verdict screen — crowding out at half: watch 7.92 talk 4.78 none 1.55 weak 0.91 strong 2.65 The verdict screen for the daily policeman says that the result depends on how fast a norm withers when nobody needs it, and the About screen gives the rates.
mechanism.js, headless, with CONFIG tagged line by line (page numbers of the paper, or “ours”); calibration.js with 43 checks; loop.js printed the walkthrough’s village (loop-log.txt) and read as a game: who grabbed, who was told off, what it did next summer.template.html draws everything on canvas in felt-tip: a wobbly line routine (every segment subdivided and jittered by a stable hash, drawn twice, the second pass faint, as a child goes over a line), a scribble fill that overshoots its outline, fish, bones, stick people with the marks the summer gave them, a sun with a face. The cards’ doodles are the same routines small. The page is a squared exercise book with a red margin in CSS; the buttons are ruled boxes with a hatch that rotates a degree and overshoots.build.py inlines the mechanism verbatim, the About page and the two fonts (Schoolbell, Short Stack) as data, and refuses to build if the embedded mechanism differs from the file.session-test.js drove the built file in headless Chromium at 380 px: the walkthrough’s village (talk, talk, watching, the occasional policeman once, watching), help opened mid-play, the eleventh summer, the verdict, About; then a second village with the daily policeman throughout to its verdict. No page errors (the hit counter’s script cannot load inside the sandbox, which is the one console line), no horizontal scroll.On Steffen’s note after playing: the start screen’s “a game from a paper” line is dropped as obvious, and the play screen and the verdict screen carry a running title, Ostrom’s Pond, in Schoolbell above everything else, so a player who arrives mid-game knows where they are. Nothing in the mechanism changed.
Steffen played it and saw the pictures and the text disagree: everyone in the row held fish while the text said the village had eaten far fewer. The cause was the fines. A grabber fined three or six fish was still drawn holding the two fish they took, and the fine came off the village’s shelf, so “we ate” fell while the hands did not. The fix follows the game’s one visual rule, nothing is counted that is not drawn: every sanction is now in kind. Under taking turns watching: a telling-off; then both fish back into the pond (the hand is empty, the pond count goes back up as it happens); then both fish back and no fishing next summer (the cross the summer after). Under a policeman: both fish to the town, drawn as a pile beside him. Deterrence in the mechanism is now two fish and a summer’s fishing, not a tariff; the watcher’s chance of seeing a grab went from a half to six in ten so that a grabber who has been told off and grabs again still expects to eat less than a sharer. One consequence: under a policeman the caught now eat nothing, so the imitation rule would have built a norm out of compliance; it no longer moves norms upward under a policeman (IMITATE_UP_UNDER_POLICE), which is the paper’s own distinction between enforced cooperation and a norm. All 43 checks pass; the same sensitivity line stands.
sweep-test.js is new: eight villages with random cards in a headless browser, and after every summer a check that the first sentence accounts for all ten villagers, every name in the text is a villager, the fish in hands equal what the text says was eaten plus the town’s take, the pond’s label and sentence equal the record, the shelf equals the running total, the eleventh summer and the verdict name the right people, and “Another village” starts clean. Eighty summers, no disagreement. Also fixed from the screenshots: notes under the figures now shrink to fit their cell and may take two lines; the town’s pile no longer runs into the sun; a villager caught by the policeman no longer wears his hat.
Steffen’s notes after playing: the aim’s last clause was hard to understand; every card is a way of putting people off taking two and the start screen should say so; the pond starts with forty fish and should say so; the feedback was lopsided (the sharers counted and underlined, the grabbers named, no count of the grabbers); and, the deep one, the player is the village meeting and would not know who took what unless somebody was at the pond. Rule adopted: every word on screen understandable to a smart eleven-year-old.
What the meeting knows is now part of the mechanism (known on every act). Everyone carries the catch home in a bucket. The meeting always learns the pond count, so it always knows how many of the ten took two; a name comes only from the village’s own watcher, who sees most villagers (six in ten), one fish or two, and reports them all, or from the policeman, who reports only whom he caught. With nobody at the pond (leave it be, have a talk) the row shows ten buckets with a question mark; the first sentence is symmetric and in the meeting’s units: “15 fish came out of the pond. 10 of us went fishing, so 5 of us took two.” The second sentence says what, if anything, was seen: “Nobody was watching, so nobody knows who”; “Ada saw 7 of us. Bo and Dot took one. Cyd took two and got a telling-off. Two more took two and got away unseen”; “The policeman caught Eli: both fish to the town. One more took two and got away.” The pond is counted only after everyone is back, so the fish leaving one at a time cannot give a grabber away. The eleventh summer is played in buckets and then revealed in full on the end screen, which the village never sees, with who never cared about the rule from the start. This is the paper’s own point about monitoring: information before punishment; and the lab’s communication treatments reveal totals, not names.
Under watching the watcher now rolls a sighting for every villager, not only grabbers, so the random stream differs from revision 2; the calibrated numbers are unchanged to the point and all 43 checks pass. The sweep gained the leak checks: no hidden name in the text, buckets drawn for the hidden, fish for the seen, and the reveal row drawing what really happened.
Text. The start screen says the pond holds forty today and fifty at most, that the fair share is one, that every card is a way of putting people off taking two, and that the buckets hide the catch. The aim: Keep the pond full and everyone fed for ten summers. Then comes an eleventh summer when nobody is watching. Will they still take one fish each? Each card now has a fourth line, what we get to find out. The paper lines on the end screen are rewritten in plain words without “norm”, “sanction”, “regime” or “crowding out” except where named as Ostrom’s word. The verdict’s small line reads “the end”.
Steffen did the sum on his first summer: forty fish, fifteen taken, twenty-five left, and then the game said thirty-nine. He expected twenty-five babies, or a cap. The code was doing what it was written to do, and that was the problem: the pond bred on a logistic curve (most babies at half full, fourteen from twenty-five fish but only nine from forty), which the screens never stated and no player could verify. Under the eleven-year-old rule that is a defect, not a detail.
The pond now has a rule in three clauses, stated on the start screen and in How to play and shown as a sum after every summer: each fish left can have one baby; the pond has food for only twelve babies a winter; the pond never holds more than fifty. Ten out and twelve in fills the pond by two a year; twelve out holds it; every fish beyond that is a fish short next summer; below twenty fish even one each leaves fewer than ten parents. The result reads “The pond had 40 fish. We took 15, so 25 were left. Over the winter 12 babies were born. Now the pond has 37 (nearly full)”, with its own sentence when room, parents or an empty pond held the number down.
What it does to the game, from the calibration run (all 45 checks pass, two of them new and about this rule): leaving it be empties the pond in about seven summers, as before; a talk now costs the pond what the promise-breakers take and leaves it about half full, where before the curve absorbed it; watching and the daily policeman fill it to fifty by summer seven; a rescue started below about twenty-five fish comes too late. The pond’s words gained “nearly full” between half full and full, since the pond starts at forty of fifty. The sweep now checks the sum on screen against the record and the rule after every summer.
Revisions 5 to 7 were a rabbit hole. Each tried to make the policeman look like the wrong answer during play, by pricing him, and each made the game harder to read without making the point better. Steffen, after a full play: nobody has a reason to go hungry in a pond that still has fish; the policeman had become a crook who takes the fish for himself; and two cards about policing in a paper about self-governance. He asked for the best version from the whole conversation, and six summers instead of ten.
So: the game is restored as it stood at revision 4 (buckets and what the meeting knows; plain words; the checkable pond rule) and cut down.
Six summers against ten moved the crowding-out rate from 0.08 to 0.12 a summer, so that six policed summers still leave a village that grabs: over 400 villages the policeman’s seventh summer has 0.9 of ten taking one, leaving it be 2.0, a talk 4.6, watching 7.3. All 41 checks pass; the sensitivity line on crowding out at half stands.
Steffen, from a playthrough: the newcomer in summer four is a complication the game does not need; and with six summers the pond could not be run down fast enough, even by leaving it be every time.
The newcomer is switched off (MIGRATION_AFTER: 0; the code and the paper’s reason for it stay in the mechanism). The pond’s food for babies goes from twelve to ten, which also makes the rule simpler to say: everyone takes one, ten out and ten in, the pond stays as it is; every fish beyond that is a fish short next summer. Over 400 villages: leaving it be empties the pond by summer five in every village; a talk ends about a quarter full and empties it in one village in four; watching and the policeman hold it level near forty. The seventh summer is unchanged: watching 7.6, a talk 4.4, leaving it be 1.9, the policeman 0.9. All checks pass; the robustness rows now include the old newcomer setting instead of excluding it.
Steffen emptied the pond in four summers, and then got a fifth summer with nothing in it. An oversight: the game ran out its six summers and the seventh whatever the pond held. Now, when a summer ends with the pond empty, the button reads “See what became of us” and leads straight to the end: “The pond is empty. After 4 summers there is nothing left to fish”, the last summer revealed (who went for two, the pond from its start to none), who never cared about the rule and that the others gave up one by one as the pond went down, and the paper’s line for the card used most, with one sentence first: the pond was fished out before the seventh summer could be played, the one thing every card was there to prevent. How to play says so. The sweep now follows a village to its early end and checks the screen.
Steffen: under taking turns watching, the row should show what everybody did, not question marks. It showed buckets for two in ten because the watcher saw eight in ten, a number I had chosen and never shown. One of us at the pond all summer sees everyone: WATCH_SEE: 1. The card says so. Nothing else changes; the first sanction is still a word, and a grabber told off once now knows for certain that a second grab costs the fish. Over 400 villages the watched village’s seventh summer moves from 7.6 to about 8 of ten taking one.