A game from a paper

Dear Household

A town wants its streets green. Every household on a street gets ten pounds and can put any of it towards saplings for its front garden: a pound, a sapling. The rest it keeps.

You are the council’s planner. You cannot make anyone plant. You write the letter: it says who will get to see what each household chose, and when. No names, ever.

Your hands will pick a letter for a street, plant the street, and bet on what the next street plants once it sees what the letter sends.

Twelve streets, six letters: get the town to plant as many saplings as you can.
A beta, built from the paper. Builder’s review only; the author finishes it.
Before you write anything

A plain street

Twenty households got the plain letter: here is ten pounds, plant what you like, keep the rest, nobody sees what you chose. Each plot is one household; each sapling is a pound it gave.

They planted 57 saplings between them. Five households planted nothing. Three planted five, half the money. One planted all ten. This is what a street does when nobody is looking; the dark mark on the bar is where it stands, and every street you write to carries that mark.

Twelve streets, six letters: get the town to plant as many saplings as you can.
Streets 1 & 2 of 12

The top street gets your letter; the bottom street sees only what the letter sends. The dark mark on each bar is a plain street, 57: a letter that works plants past it. Touch a letter to see who will see whom, then plant.

Touch a letter.
The town, twelve streets later

your town
plain letters
best letter

What the paper found

How to play

Dear Household

Twelve streets, six letters: get the town to plant as many saplings as you can.

The streets. Every household on a street gets ten pounds and may put any of it towards saplings for its front garden, keeping the rest. A plot is one household; a sapling is a pound it gave. A plain street, where nobody is told anything and nobody sees anything, plants 57 saplings. Green grows only where someone chose it.

The letter. Streets come in pairs. The top street gets your letter, which tells each household who will see its pledge, and when. The bottom street sees only what the letter sends, before it decides. Touch an envelope and the picture shows who will see whom: an eye on a plot that is told it will be seen, cards and arrows on the road for what travels, a slot on a plot that receives something. The six letters are: the plain letter (nobody sees anything); the sheet (all twenty pledges to every household on the next street, before they decide); the spokesperson (one household’s pledge to all twenty on the next street, before; the other nineteen told nothing); pass it on (each pledge to one household on the next street, before); afterwards (each pledge to one household on the next street, after they have decided); the clipboard (all twenty pledges to one household on the next street, before). No letter carries a name.

Plant this street. The top street fills, plot by plot, and its bar and number show how many saplings it planted against the plain mark at 57. The game then says what the letter told the households, what they planted, and what the paper found about that letter.

The bet. When the letter sends something to the next street before it decides, you say how many saplings the next street will plant. The slider starts at the plain street’s 57; the marigold marker on the bottom bar shows your bet. Send, and the bottom street fills. The game says what they saw, what they planted, what you expected, and what the paper found. When nothing reaches the next street in time, there is no bet, and it plants like a plain street.

Next streets. Six pairs, one letter each; you may send the same letter again. Nothing carries over from one pair to the next. After the sixth pair the town is counted: your saplings against a town of plain letters and against the best letter on every street, and the paper’s word for how you bet.

Dear Household

About

Built from Jayant Vivek Ganguli and Friederike Mengel, “Social influence and carbon dioxide mitigation”, Journal of Public Economics 253 (2026), 105558. DOI 10.1016/j.jpubeco.2025.105558, open access. Preregistered as AEARCTR-0010768. Game built from this paper; the paper is the authority where the two disagree.

What the paper found

Four thousand seven hundred adults in the United Kingdom, online, each given ten pounds and asked how much of it to spend on planting trees to offset carbon, keeping the rest. Nobody saw anybody’s name. Before choosing, some were told that one other person, in a second round, would see their choice before making their own; some that the other person would see it only afterwards; some nothing. The ones who knew one person would see their choice before deciding gave about a sixth more. The ones whose choice would be seen afterwards gave what the ones told nothing gave. So it is not being watched that moves people; it is the chance that their choice shapes someone else’s.

Then the authors tried to scale it. Told that twenty people would see their choice, each seeing only theirs, people gave no more than when one would. Told that one person would see their choice among nineteen others’, they gave as if nobody would see it. Told that twenty people would each see theirs among nineteen others’, only slightly more than that. Influence, in their words, does not easily scale up: more eyes on one choice add nothing, and a choice that is one of many is as good as unseen.

The second round showed how far the influence actually went: a person who saw a choice of ten pounds gave about seventy-five pence more than one who saw a choice of nothing, seven or eight pence for each pound seen. Asked beforehand to guess, the first round expected five or six times that. Only about one in ten guessed within a tenth of the truth; most overestimated their influence, one in five by a lot; one in six underestimated it. The authors also note that people underestimate what others give when nobody is looking. In all, the participants’ choices paid for offsetting about two thousand tonnes of carbon dioxide.

What the game does

The experiment becomes a town. A street of twenty households gets your letter, which is one of the paper’s arrangements with a plain name: the plain letter, the sheet, the spokesperson, pass it on, afterwards, the clipboard. The street plants; each pound is a sapling in a plot. The next street sees what the letter sends and plants in turn, after you have bet how many saplings it will plant, which is the paper’s own question about influence asked about a street instead of a stranger. Six pairs, then the town is counted against a town of plain letters and against the best letter everywhere, and the game says, in the paper’s four words, how you bet.

The mechanism, and what it claims

Kind: reconstructed. The game runs a data-generating process built from the paper’s estimates: each household on the lettered street pledges, on average, the paper’s plain amount plus the paper’s effect for the arrangement it was told; each household on the next street that sees a pledge before deciding adds the paper’s seven and a half pence for every pound by which what it saw differs from a plain pledge. The plots show whole saplings drawn from the shape of the paper’s choices, with its lumps at nothing, at five and at ten, so that every street adds up to its average.

The game demonstrates the paper’s estimates as a street shows them: the sixth more for a pledge one person will see before deciding; nothing for a pledge seen afterwards; nothing more for twenty viewers than for one; nothing for a pledge that is one of twenty; the slight rise when twenty see twenty; and the small copy, seven pence in the pound, by the one who sees.

The game assumes everything the paper did not run. The paper gave every observed person twenty observers by recruiting twenty people each; the game’s spokesperson is one household on a street seen by the twenty on the next, with nineteen told nothing, because in a town attention is a fixed resource. The effect of twenty viewers is played as exactly nothing where the paper’s point estimate is small, positive and not distinguishable from nothing; so is being seen afterwards, and being one of twenty seen by one person. The copy on the next street is centred on the plain pledge, so that a household which sees an ordinary pledge plants the ordinary amount; the paper’s own numbers put it a few pence lower. A household on the next street only sees; it is never also told who will see it, so the two effects are never added in one person. Each street is the paper’s average street with no chance in it: a real street of twenty would swing by fifteen saplings either way, more than the effect, which is why the paper needed a thousand people to see it. The six letters as real-life arrangements, the pairs of streets, the saplings, the bet’s range and its cutoffs are ours.

The paper establishes the average effect of each of its arrangements on what people gave, in the United Kingdom in 2023 and 2024, with standard errors; that the first-round effect is influence and not observability; that the effect of twenty viewers is small and not significant, with limited power; that being one of twenty seen removes the effect; the response of the one who sees, per pound seen; and that most people overestimate that response. It does not establish what the same letters would do in a street where names are known, what a town would do over years, or whether the influence propagates beyond one step.

Where the discovery comes from. That pass it on beats every other letter follows from the paper’s cells: twenty households each told one person will see them, against one household, or none, or households told they are one of twenty. The calibration check shows it survives the paper’s point estimates in place of the zeros, a different centring of the copy, twice the copy, no copy at all, the sheet copied at its mode rather than its mean, and the bottom of the main effect’s confidence interval.

The numbers

Every number sits in the config block at the top of the mechanism, tagged with its table. The plain pledge and the effect of one viewer are Table 2; the effects of twenty viewers, of being one of twenty, and of both, Table 3; the copy per pound seen, Table 4; the beliefs, the multiplier and the types, the text of section 3.3, footnote 14 and Figure 1. The shape of pledges is read off the histogram beside Table 2 and adjusted to the reported means; it decides only how the plots look, never what a street adds up to. On screen the only numbers are saplings.

What is left out

The observed-afterwards arrangements with twenty viewers or twenty seen, reported only in Figure 1; who gives more (those who have thought about the climate, worry, feel responsible, or believe they inspire others) and the fact that the letter’s effect is the same for all of them; the underestimation of what others give; the checks on ceiling effects and fast responders; the welfare discussion; and the heterogeneity in beliefs by class, origin and religion.

Credits

Built with Claude under the paper-to-game skill. Type: Fredoka (Milena Brandão, Hafontia), SIL Open Font License, embedded. Source, design memo and calibration check: github.com/steffenhuck/paper-to-game.