Trust Calibration

Trust calibration is the dynamical process by which agents update their assessments of another's reliability based on accumulated evidence from interaction outcomes. It is not a Bayesian computation performed in isolation; it is a network-coupled phenomenon in which each agent's calibration is shaped by the calibration of their neighbors. When your friends distrust a source, your own distrust amplifies even without direct experience — a mechanism that can produce information cascades of misplaced trust or distrust.

The process connects epistemic vigilance — the cognitive mechanisms for evaluating testimony — to structural causation: the network topology determines which evidence reaches which agents, and therefore whose calibration is accurate and whose is systematically distorted. In tightly clustered trust networks, calibration errors can persist for long periods because dissenting signals never cross community boundaries. The study of trust calibration thus requires not merely psychology but reputational dynamics — the formal analysis of how reputation propagates, decays, and saturates in social networks.

The calibration of trust is not a rational update; it is a social process that only sometimes produces rational outcomes.