Trust in predictive maintenance: effects of algorithm reliability and downtime risk

This study investigated how recall-defined algorithm reliability and downtime risk, operationalised as downtime loss, influence trust through two experiments in which student participants completed simulated predictive maintenance (PdM) tasks under deterministic and uncertain risk conditions. Results showed that higher reliability increased trust attitude and shortened decision time across both experiments. It also reduced recommendation rejection under deterministic risk and in the exploratory analysis under uncertain risk. However, the effect of downtime risk varied: under deterministic conditions, higher downtime risk reduced trust behaviour without significantly affecting trust attitude; under uncertain conditions, the variance-based nominal risk manipulation did not reliably differentiate perceived risk and did not significantly affect trust-related outcomes, whereas experienced downtime losses were associated with lower trust attitude. These findings highlight the importance of improving PdM-relevant detection performance and effectively communicating risk and outcome information to support informed maintenance decisions and effective use of PdM decision-support systems.