Designing institutions for low-trust environments

Most writing about institutional trust is written from places that still have some. It assumes a baseline of goodwill: that people will extend an institution the benefit of the doubt, and the institution’s job is not to squander it.

Much of the world does not work that way — and designing for that reality turns out to teach something important about designing for everywhere.

Distrust as the rational default

The numbers are stark and durable. Latinobarómetro, which has surveyed Latin American publics for three decades, consistently finds trust in political parties, congresses, and courts at some of the lowest levels ever recorded — trust in parties has run near or below 20% across the region for years, and satisfaction with how democracy works has hovered similarly low (Latinobarómetro). The OECD’s surveys of trust in public institutions show the problem is not confined to any one region: in many member countries, fewer than half of citizens say they trust their national government (OECD Trust Survey).

It is tempting to read this as cynicism. It is more accurate to read it as experience. Where institutions have repeatedly failed people — corruption discovered after the fact, promises broken without consequence, decisions made in rooms no one can see into — withholding trust is not a bias. It is learning.

What low-trust environments teach

Here is the design insight: an institution serving a low-trust public cannot ask to be believed. It has to be checkable.

That distinction changes what you build. A high-trust institution can say “we reviewed this carefully” and be done. A checkable institution publishes what it reviewed, shows who signed, and leaves a door open for challenge — because it knows the reasonable reader will not take its word, and it designs so they don’t have to.

The strongest anticorruption and transparency work of the past two decades follows exactly this pattern. Open-contracting reforms didn’t ask citizens to trust procurement officials more; they published the contracts. Budget-transparency movements didn’t ask for faith in treasuries; they made the ledgers legible. The trust, where it came, came after the verification was possible — as a consequence, not a precondition.

Why this matters for AI now

AI-assisted decisions are arriving into this landscape, and the default design of AI systems is precisely the one low-trust environments have learned to reject: an opaque process, an unnamed decider, a confident answer, and an implicit trust us.

In Bogotá or Buenos Aires — or, increasingly, in Brussels and Washington — that posture fails on contact. A public that has learned to distrust unexplained authority will not make an exception for an algorithm, and should not.

So we build to the opposite posture, and we state it as a rule: assume the reader does not trust you, and make that not matter. Every recommendation carries evidence that can be opened. Every consequential output names a person who answers for it. Every answer can be challenged through a real channel. None of this requires the user’s goodwill. It requires only that they can check — which is the one thing a rational skeptic will do.

The general lesson

Institutions designed for low-trust environments are not a special case of institutional design. They are the honest case. The high-trust world’s goodwill was always partly a subsidy, and it is visibly running out everywhere.

Terraced hillsides are built for the hard case — for the steep slope and the heavy rain — and they hold for centuries. Institutions are the same. Build for the skeptical reader in the low-trust environment, and you build something everyone else can rely on too.

Sources

• Corporación Latinobarómetro. Informe Latinobarómetro (annual regional survey). latinobarometro.org
• OECD. Survey on Drivers of Trust in Public Institutions. oecd.org — trust in government
• Open Contracting Partnership — evidence and case studies on procurement transparency. open-contracting.org