74. The Rule of 5 🔗

There’s a theory that the ability to have painful conversations in large groups is some sort of valuable and advanced skill that we must practice. I submit that this skill is psychological science fiction. No group larger than 5 people has ever had such a conversation.

Above 5 people, any overture to start such a conversation, no matter how hard you try to be kind, will be an accusation of moral crimes leveled against an abstract archetype that will result in polarized derp based identifying with/against that archetype.

Try not to identify strongly with any archetype of n > 5, even if you can’t escape the consequences of such identification on the part of others. Try to base any response on an identification of the counter-party with a < 5 group. Small enough to name all members.

This thread prompted by yet another profoundly futile outrage cycle triggered by yet another attempt to spark such a conversation. I won’t link to it. But it’s not enough to starve such things of attention. You need alternative ways to talk about those things.

I propose a “Rule of 5” model.

Stop criticizing or holding accountable large abstract classes of people. And stop trying to make an “example” of members of the class, no matter how egregious.

Instead pick on 5 named people, living or dead.

5 is large enough to establish a structural pattern worth acting on, and force you to build a somewhat general moral case, but not so small that you’re unfairly picking on one “example.”

The rule works for most kinds of passive complicity in real but nebulous bad things.

Pick on individuals only when they are structurally special. Like Epstein or Trump, catalyzing a lot of bad from a uniquely privileged structural position. Not just “public figures.”

Only when you can’t find 4 other examples in their close milieu.

I think even billionaires deserve the pattern-of-5 model. There’s enough of them that if you think 1 billionaire is doing a bad thing you should be able to find 4 more.

This is not to protect them, but to protect you from the temptations of careless and futile moral jousting.

A good example is the USC admissions scandal. They took down enough at once that the pattern was clear. Yes they probably took down the vulnerable ones and worse offenders likely got away, but it was a decently surgical strike. As good as it can get probably.

Buffet’s heuristic of “praise specifically, criticize generally” is basically wrong in 2021. It only works in narrow, homogeneous, contexts that don’t contain their own externalities. A kind of context that’s increasingly hard to sustain without doing sketchy kinds of gatekeeping

If your context has sufficient scope and heterogeneity, both praise and criticism should be small patterns by default. Hence Rule of 5.

Praise/criticize in named groups of 5 by default
Individually by exception
Archetypally never

If something good is happening, chances are more than one person is riding a systemic serendipity. Picking out 5 lets you praise individuals a bit but also flag the pattern, inviting others to ride it, and counter-programming the temptation to feed attribution bias.

It’s fine to traffic in archetypes or even stereotypes when you sense the conversation is not “difficult” and within the scope of good-natured jokes/roasting that will be received as such. This means enough mutual trust and small enough power differential.

Been thinking along these lines for a few years, but I was stuck on 1:1. That’s an ideal in some ways. When you engage 1:1 it’s hard to truly dehumanize. You’re forced to see the other person as a human and if not, have your inhumanity witnessed. But it doesn’t scale.

The rule of 5 scales, works at a distance, works on large orgs, etc. Forces due diligence, generates proof of systemic problems in the process of formulation etc.

If the rule in a court of law is “innocent until proven guilty” the rule in a court of public opinion is “guilty until proven innocent” and there’s no good middle ground. Rule of 5 is not as sloppy and type-1-error-prone as mob law, not as type-2-error-prone as court law.

No point decrying mob law/court of public opinion. It’s a permanent feature of the species. No point trying to make courtroom law more expansive. Thar be abuses. But we can invent a third type.

Juries have guardrails unstructured conversations don’t, so they can scale a bit. tweet^[1]

Just occurred to me that the NYT practices a perverse inversion of this trend, when there are ~3-5 named individuals on the provocation side as in “these 3-5 people claim iPad use causes racism.” A specious “difficult conversation” premise propped up by bearing anecdotal witness.

Another way to remember this heuristic: to a first approximation, the largest group that can do effective unstructured conflict resolution is a nuclear family. And it really sucks even at that scale. Above that you’re just picking a fight that will reproduce bad patterns.

For eg if a group A is trying to fight a group B for change in an institution, and A+B >5, then whatever the outcome, it will harden and further reify the most visible archetypal divide between A and B. You’ll “rotate” the problem rather than solve it.

The synthetic link between “5 is the useful limit” and “name 5 people” is perhaps not obvious. It’s just that that that’s what our social brain can really handle. Some sort of speculative cousin of Dunbar number.

Above/below you get overfit/underfit

This is part of my thesis that a controlled dark age is probably for the best now. The world needs a good night’s sleep. Been “enlightened” too long. Going < 5 = going cozy = going dark. Public spaces only for light entertainment. Take difficult stuff backstage. Breakout sessions