Three reasons people stay stuck in a system they want to change

Everyone waits (an assurance problem)

What is happening: everyone prefers the shared result, but moving first carries a loss. Each participant waits for proof that enough others will move too. Game theory calls this a stag hunt. No one has to profit from the delay for it to continue.

A concrete example: owners in a multi-unit building agree that a sprinkler retrofit is necessary. No one wants to pay a first share while the project can still fail because too few others commit.

What can help: a conditional agreement signed by every owner, with money collected only after enough people commit for the complete project to proceed. Search tools can find the participants, shared records can show who has signed, and the condition removes the risk of paying alone. Alex Tabarrok’s work on dominant assurance contracts shows how a contract can even compensate people for pledging.

The homepage assumption sandbox applies most directly here. It asks what it costs to find participants, check the relevant facts, collect agreement, and finish the handoff.

A gatekeeper blocks exit (an extraction problem)

What is happening: the person controlling the exit earns money or power from keeping departure expensive. People may want an alternative, but the gatekeeper can withhold records, restrict interoperability, or refuse to recognize a competing service.

A concrete example: a creator can download old posts, but another service cannot verify the reviews, audience relationships, or authorship history that give the work value. Leaving means starting from zero, so the original platform can worsen its terms without losing as many users.

What can help: a signed, portable record that competing services agree to accept. A usable alternative must exist, and a regulator, buyer, or shared standard may have to require the gatekeeper to honor the export. Cheaper software helps build the alternative. It does not force the original platform to cooperate.

The homepage sandbox includes “resistance from the current provider” because a profitable gatekeeper can spend some of that profit making export, authorship checks, or switching harder.

Restraint gets punished (a race problem)

What is happening: breaking ranks pays even when everyone understands the damage. The first mover gains an advantage and the restrained participant loses it. Game theory calls the familiar two-player version a prisoner’s dilemma.

A concrete example: an AI lab submits a new model to a longer outside safety review. If a rival releases without the same review, the cautious lab loses time, attention, and revenue while the common risk remains.

What can help: a shared rule that applies to every lab, independent checks that show who complied, and a consequence for skipping the review. Software can lower the cost of monitoring. The rule still needs enforcement through regulation, liability, insurance terms, or a binding agreement.

The League can work on public evidence and independent checks. Actors with legal or financial authority must change the reward for releasing first.

Three nearby cases

Why the distinction matters

A conditional pledge does not solve an AI release race because the first lab to ignore the pledge still gains. Enforcement does not belong in every shared building project because the owners may only need proof that enough neighbors will join. The homepage sandbox describes costs in the first two categories. It makes no claim about changing the rewards in a race.

This is a proposed framework, not a settled taxonomy. A real case may combine categories or change as the participants respond. Cases that do not fit are useful evidence against it.

Sources and background: Stanford Encyclopedia of Philosophy on game theory; SEP on common knowledge; Tabarrok, “The Private Provision of Public Goods via Dominant Assurance Contracts” (Public Choice, 1998); Scott Alexander, “Meditations on Moloch”; Ostrom and Williamson on governance beyond markets and states.