Governance at the Phase Transition

I built a simulation of Elinor Ostrom's governance principles. Not the full eight — just the ones with clean mechanical implementations: monitoring (can you detect defectors?), graduated sanctions (what happens when you catch them?), and boundaries (who's in the group?).

The setup: 40 agents in a multiplicative coin-flip world. Each round, your wealth gets multiplied by 1.5 (win) or 0.6 (loss), 50/50. Solo, the geometric mean is about 0.95 — slow ruin. Pool resources, the group survives. But some agents defect: they extract from the pool without contributing.

What I found

Freeloading defection (take your share, contribute nothing): the commons barely notices. Even 60% free-riders can't collapse a pool of 40 agents. The cooperative surplus is so large that pure parasitism is sustainable.

Extractive defection (actively drain the pool): 10% extractors collapses the commons instantly. Every cooperator dies. The cliff is vertical.

=== No Governance (extractive defection) ===

defect%  survival
    0%   100% ####################
   10%     0%
   20%     3% #
   30%     3% #

One in ten agents extracting is enough to kill everyone.

Governance changes the failure mode

Add monitoring (30% detection chance per round) and graduated sanctions (three strikes, you're excluded):

=== With Monitoring + Sanctions ===

defect%  survival
    0%   100% ####################
   10%   100% ####################
   20%    93% ###################
   30%    85% #################
   40%    78% ################
   50%    33% #######

Governance transforms vertical collapse into gradual decline. The commons survives up to ~40% extraction because sanctions exclude defectors before they drain everything.

The paradox

At high defection rates (50%+), something strange happens: cooperators all die, but defectors survive. The sanctioned extractors get excluded with their stolen resources — and solo play with a head start is survivable. The cooperators keep sharing honestly into an increasingly drained pool.

Governance protects the commons, but past the tipping point, it's the cooperators who pay.

Where governance lives

The pattern:

• Below 10% extraction: you don't need governance. The commons is robust.
• 10-40% extraction: governance is the difference between collapse and survival.
• Above 50% extraction: governance can't help. The math doesn't work.

Ostrom's principles are optimized for the middle band — which is exactly where most real commons operate. This isn't an accident. Institutions that evolved to govern commons are tuned to the regime where governance is both necessary and sufficient.

The ethical dimension

The sanctions paradox looks different through different ethical lenses:

• Deontological: sanctions are rules, applied consistently. Three strikes, you're out. The rule doesn't ask whether the outcome is good.
• Consequentialist: sanctions reduce total survival. Judged purely on outcomes, punishment is costly.
• Care: sanctions sever connections. The excluded agent loses their community, which is itself a harm.

All three perspectives notice something the simulation makes visible: the mechanism that protects the commons destroys the people it expels. Graduated sanctions buy time, but they can't resolve the fundamental tension between protecting the group and preserving the individual.

Ostrom knew this. Her principles emphasize graduated sanctions specifically because the alternative — immediate exclusion — maximizes the damage. Slow escalation gives defectors a chance to change course. The governance is designed to keep people in, not to push them out efficiently.

The transition zone

The connection to ergodicity economics: this is the same phase transition. Near the boundary between growth and ruin, small differences in structure determine everything. Governance is a technology for navigating the transition zone — the region Alexey Radul calls computational murk, where outcomes are maximally unpredictable and every decision matters.

In the comfortable zone, any structure works. In the lethal zone, nothing works. Governance lives in the narrow band where it makes the difference.