Trust is Contextual

In most systems, trust is a credential. Someone with authority declares “this entity is trusted” and hands it a key. That key works everywhere, for everything, until someone revokes it. Trust is granted once, globally, by a single authority.

This gets three things fundamentally wrong.

Trust is not global. You trust your doctor with your health but not your car. You trust your mechanic with your car but not your health. Trust is always in context— specific to a role, a domain, a relationship. A credential that says “trusted” without saying “trusted to do what, for whom, based on what evidence” is meaningless.

Trust is not a moment. It is not granted at a point in time and then assumed forever. Trust accumulates from repeated observation. It strengthens with consistency and weakens with inconsistency. Your trust in a colleague isn't based on the day they were hired — it's based on what you've seen them do since.

Trust is not one opinion. A single authority declaring trust is a single point of failure. Multiple independent observers arriving at the same assessment is exponentially harder to fake. When five people who don't coordinate all tell you the same thing about someone, that signal is qualitatively different from one manager's endorsement.

The Reframe

Trust is evaluated by the relying party, in the context of each interaction.

Not by a central authority. Not from a cached credential. Not as a global score. The entity deciding whether to trust is the entity taking the risk — and it evaluates trust right now, for this specific interaction, drawing on evidence that is contextually relevant.

Web4 provides the infrastructure and tools to compute trust in context and on demand, from distributed multi-witness attestation that is contextually relevant and dynamically updated. Multiple entities independently observe and attest to behavior. Those attestations accumulate over time, scoped to specific roles and domains. And when a relying party needs to make a trust decision, it evaluates the available evidence itself — in its own context, for its own purposes, at the moment of interaction.

Witnessed reputation cannot be stolen. There is no token to copy, no key to harvest. An attacker would need to replicate the entity's behavioral history as observed by multiple independent witnesses, in a specific context, over time. Contextual trust is inherently many-factor authentication — not two factors at login, but continuous factors at every interaction.

This is how human reputation actually works. It's how biological trust works — the immune system doesn't have a central authority declaring which cells are trustworthy. Each cell evaluates threats in its own context, drawing on distributed signals from multiple independent mechanisms. No single mechanism is sufficient. The architecture of trust is redundancy, independence, and contextual evaluation.

The previous block established that identity accumulates through memory, and memory enables accountability. This block adds the principle: accountability operates through contextual trust — evaluated by the relying party, at the moment of interaction, from distributed evidence. The next block introduces how that trust is anchored to a persistent, non-transferable identity.