Interpreting risk drift in construction projects: a fuzzy–Bayesian decision framework for weak signal–based governance

Construction project risks rarely materialize through sudden threshold breaches; instead, they tend to evolve gradually as weak and ambiguous signals accumulate under persistent delivery pressure. Conventional construction risk management systems, which rely on discrete indicators and fixed escalation rules, remain poorly suited to recognizing this form of gradual confidence erosion, commonly described as risk drift. This study develops a decision-oriented fuzzy–Bayesian framework to operationalize risk drift as an evolving belief about project safety rather than as a static or predictive risk state. Fuzzy logic is used to translate heterogeneous monitoring indicators into interpretable linguistic concern levels, preserving ambiguity inherent in early warning signals, while Bayesian updating accumulates these concerns over time to represent progressive belief change. The framework is illustrated using a publicly available integrated construction monitoring dataset to demonstrate how risk drift can emerge even when individual indicators remain within acceptable ranges. The results show that sustained marginal conditions can erode confidence in safe operation without triggering conventional alarms, highlighting limitations of threshold-based risk logic. By reframing risk as a belief-based governance construct, the proposed approach supports earlier, proportionate, and precautionary managerial intervention using existing project data, without reliance on complex prediction models or rigid escalation thresholds.