A one-sheet on posture toward the people around you
The principle, and its known failure
Jon Postel wrote it for computers talking to each other: send strictly correct output, accept anything you can reasonably interpret. It let a fragile early internet work despite everyone’s software being slightly wrong.
Engineers later turned against it. Broken implementations survived because everyone quietly worked around them. Malformed input became an attack surface. The real standard drifted from the written spec toward “whatever people actually send.” Later standards walked it back.
Both halves transfer, but not symmetrically. In protocols, liberal acceptance causes drift — the spec rots slowly. In people, it causes adaptation — the other party notices and adjusts. Which is the thing to hold onto: charity here isn’t only a private virtue. It’s an intervention in a system that reacts to it.
Three tiers, not two
Form — be liberal. The blunt email. The typo. The question asked badly. The friend who is bad at expressing warmth. The tone that landed wrong on a day you were already raw. Charitable reading here costs you nothing and buys enormous goodwill.
Substance — be conservative. Claims, requests, commitments, and resource flows get evaluated on their merits. Forgiving how something was said is not the same as accepting what was said.
Pattern — be strict, and early. This is the tier most people miss, and it’s where form quietly becomes substance. One blunt message is noise. A repeated pattern of bluntness aimed at you is information — not about their communication style, but about how much care they think you’re worth.
Form and substance are a first-pass parser, not a permanent taxonomy. Repetition promotes form to substance, and once promoted it doesn’t get the charity anymore.
Most people collapse the tiers in one of two directions: strict on form, so everything reads as an attack; or liberal on substance, so anything delivered pleasantly gets through. Both are errors in the same variable — failing to notice which layer is being evaluated.
The sending half is about noise, not volume
“Conservative in what you send” gets misread as send less, which turns into coldness, withholding, and the slow starvation of relationships that run on small imperfect gestures. Warmth requires half-formed bids — the tentative invitation, the slightly awkward over-disclosure. Those are you asking someone else to be liberal about form, which is exactly the trade this document endorses.
The correct reading is low ambiguity, not low bandwidth. Don’t make people decode you.
Ambiguity is a tax. You levy it on everyone downstream while paying almost none of it yourself — which is precisely why so few people control it.
Layer it, don’t calibrate it
The internet didn’t solve this by becoming strict. It became layered: still promiscuous at the transport layer, because in an open network you can’t choose who contacts you — but heavily authenticated above it.
The social version:
- First contact: generous. Cheap, reversible, and the cost of being wrong is small.
- Trust extension: strict. Access to your time, money, reputation, and vulnerability is granted deliberately and in increments.
Read form charitably. Don’t extend the same charity to substance or pattern.
Reading a network without paying for the information
Being lenient does reveal who exploits leniency — but it’s an active test, and running it changes the result. Leniency doesn’t only expose the cost of defecting against you. It lowers it. You aren’t sampling a fixed distribution of behavior; you’re shifting one. Prefer tests you can run by watching.
Watch the first small violation. Healthy networks self-correct early, cheaply, and from the periphery — an ordinary member says something, it’s absorbed, everyone moves on. Unhealthy ones either ignore it or escalate straight to authority.
Watch how someone with no leverage gets treated. The intern, the new person, the one who can’t retaliate. Status compresses information; powerlessness decompresses it, because nobody performs for an audience that can’t affect them. Highest signal available, and it costs nothing.
Watch what happens to the person who raises a problem. If complaint reliably costs more than the thing complained about, you are looking at a network where nothing will be fixed.
What actually turns a network uncooperative
Some people in every large group are dispositionally exploitative. The exact proportion is secondary. Cooperation is mostly structural, not characterological — most defection is ordinary people responding to incentives.
What matters is distribution: whether such people are scattered through the network or concentrated in its gatekeeping and enforcement roles. A group absorbs a lot of scattered predation. It cannot absorb predation sitting on its immune system, because then every good-faith complaint feeds the thing meant to correct it.
Four conditions do most of the work:
- Encounters stop being repeated and identifiable. Reputation is the cheap enforcement mechanism. Scale and anonymity destroy it, and nothing equally cheap replaces it.
- The pie stops growing — or seems to. Cooperative norms are partly a luxury of expansion. Perceived stagnation is enough; the pie doesn’t actually have to stop.
- Detecting defection costs more than defecting. When tolerance is diffuse but strictness falls on whoever enforces it, nobody enforces, and the effective norm ratchets toward whatever people actually do.
- Repair gets replaced by escalation. Once the first response to error is punitive rather than corrective, everyone starts pre-defending — which is behaviorally identical to defecting.
The governing rule
Everything above reduces to this:
Generosity builds cooperation when it sits on top of enforcement. It destroys cooperation when it substitutes for enforcement.
Where repeat play is likely and correction is still cheap, charity is productive — it’s the thing that makes the relationship worth having. Where there’s no repeat play, or where nobody can afford to say “that wasn’t okay,” charity isn’t kindness. It’s an unpriced subsidy, and subsidies attract takers.
The part that dissolves the question
You are rarely in a network. You’re on a set of edges with different histories, stakes, and exit costs. Your family, your employer, your group chat, and the platform you argue on are four different games with unrelated payoff structures, and the useful resolution is almost always the individual relationship, not the graph.
So the question isn’t “am I in a cooperative network.” It’s “on this edge, under these conditions, what’s the cost of being wrong in each direction?”
Where being wrong is cheap and reversible, be generous — the generosity is usually what makes the edge cooperative in the first place. Where it’s expensive and hard to undo, be strict, and be strict early. The cheapest correction is always the first one.
cafebedouin.org 