The network effect of trust

Suppose one firm decides to do the disciplined thing. It keeps a real account of its consequential decisions — the evidence that was actually in front of each one, the rules that were active at the time, enough state to replay the decision and see whether it holds. This is a genuine improvement, and it pays. The firm can answer its own questions after the fact, defend itself when challenged, and find its own mistakes before someone else does. But the benefit is private and it is bounded. The firm has built a better filing cabinet. What it has not built, and cannot build alone, is anything you would call infrastructure.

The difference between a good private practice and a piece of infrastructure is not the quality of the artifact. It is whether anyone else is obliged to recognize it. A record that only its author knows how to read is a diary. A record in a form that a regulator, an insurer, a counterparty, and a court all already know how to read is something else — it is a common instrument, and its usefulness has almost nothing to do with the firm that produced this particular one. The value has migrated out of the object and into the agreement about what the object is.

That migration is the whole subject here. Accountability infrastructure exhibits network effects, in the strict economic sense: the value of holding a decision record to any one party rises with the number of other parties who hold records in the same form and honor the same expectations about what such a record must contain and how it can be verified. The record is only as strong as the standard it conforms to, and the standard is only as valuable as the crowd that has adopted it.

From curiosity to infrastructure

Consider what has to be true before a decision record does any work outside the firm that made it. Another party has to be able to open it, understand its structure, know which fields carry the evidence and which carry the rules, and check the provenance and the replay without a phone call and a bespoke integration. Every one of those requirements is a requirement about shared expectations, not about the record. You can produce a technically perfect account and, if no one else recognizes its shape, it accomplishes nothing beyond your own walls. Conversely, a merely adequate record in a universally recognized form travels everywhere.

This is the oldest lesson in the economics of standards, and the history of infrastructure is mostly the history of it being learned the hard way. A railway is worthless track until the gauge is agreed; the moment a continent settles on one width, every mile of track becomes reachable from every other, and the value of the network is not additive but combinatorial. The shipping container did nothing on its own — a steel box is a steel box — but a standard box, sized so that any crane, any ship, any truck bed, and any rail flatcar in the world can carry it without being asked, rebuilt global trade. Accounting standards let a lender in one country read the books of a borrower in another without auditing the language from scratch. A file format lets a document written on one machine open, unchanged, on a machine its author will never see. In each case the artifact is unremarkable and the agreement is the asset.

An admissible record is the same kind of object. A Decision Receipt — a record built to be examined, carrying its own evidence, its own frozen rules, and enough to replay the decision — is a real thing whether or not anyone else keeps one. But it becomes a rail gauge only when enough parties keep the same shape that keeping a different one starts to feel like speaking a private language at a public meeting.

The first admissible record is a novelty; the ten-thousandth, all in the same shape, is an institution — and by then no one can afford to speak a different language.

Value that compounds with adoption

Once a common form exists, each new adopter makes it more valuable to every existing one, and the mechanism is worth spelling out because it is not vague goodwill; it is specific and it accumulates. When two parties keep records in the same shape, either can check the other's without building anything — the integration was paid for once, by the standard, rather than pairwise by every relationship. Add a third party and you have not added one connection but several; the checkable relationships grow faster than the membership, which is exactly the signature of a network good. The cost of verification, which is the real expense that keeps accountability underprovided, falls for everyone each time the circle widens.

The compounding does not stop at the parties who keep the records. A regulator who can name a known artifact — produce the record, in the standard form — has a far easier job than one who must specify from scratch, per firm, what an acceptable account looks like. An insurer who can price against a common bar can underwrite decision-making as a category, because the thing it is pricing has a stable definition across the whole book rather than a hundred private ones. A court that has seen the familiar form before does not have to be taught, case by case, why this kind of evidence deserves standing. Every one of those actors is, in effect, another adopter, and each one that comes to recognize the form raises the cost to any individual party of not producing it.

This is why standards tip, and why, once tipped, they are so hard to dislodge. Adoption lowers the cost and raises the payoff of further adoption, so the process feeds itself past a threshold and then runs to near-completion on its own. And the same feedback that builds the standard defends it: displacing an entrenched standard means persuading everyone to switch at once, because the first to leave loses the network and gains nothing, so no one goes first. A rail gauge chosen for indifferent reasons outlives every argument for a better one. The lock-in is not a conspiracy; it is arithmetic. The switching cost is the whole installed base, and no single party can pay it.

The prize is the standard

All of which points to a conclusion that anyone building in this space should sit with honestly. Accountability will not consolidate around a scatter of proprietary silos, each firm hoarding a private format legible only to itself, for the same reason trade did not consolidate around a thousand incompatible box sizes: the silo forfeits precisely the network value that makes the record worth keeping. It will consolidate around standards. And that means the strategic high ground is not any particular record, nor even the ability to produce records well. It is the definition of the common record itself. The party that sets what an admissible account must contain and how it may be verified shapes the market that grows on top of it, the way the party that fixed the gauge fixed the shape of every railway that followed.

That is a real prize, and it would be dishonest to pretend it is an unmixed good. A standard is a chokepoint as much as a commons. The same lock-in that makes a good standard durable makes a bad one durable too, and standards can ossify — freezing yesterday's assumptions into a form no one can afford to revise — or entrench whoever controls them, turning a shared instrument into a private tollgate. A closed standard, verifiable only by its owner or licensed only to its friends, would reproduce inside the accountability layer the exact opacity that accountability exists to dispel. It would be an unaccountable arbiter of accountability, which is worse than none.

The way through is not to avoid the standard but to insist on the right kind. An open standard — one whose form is public, whose records anyone can verify without permission, whose replay depends on no privileged access — serves the ecosystem, because its value flows to every adopter rather than being taxed by an owner. A captured one serves the incumbent who holds the keys. Both enjoy the network effect; only one deserves it. The prize is worth pursuing, and worth pursuing precisely so that it is won in the open, by a form that earns its ubiquity by being checkable rather than by being enclosed. The market for trust will settle on a common shape whether we design for it or not. The only question left open is whether the shape it settles on is one that anyone can hold to account, or one that no one can.

— Dispatches · Summit Cognitive