Automation Reliability Impairs Evidence Accumulation Efficiency: Computational Modeling of Monitoring Under Time Pressure

Automated decision aids are increasingly deployed in safety-critical domains, yet their inevitable imperfections require operators to continuously monitor and independently verify system recommendations. Prior research has established that time pressure and low automation reliability impair monitoring performance, but the underlying cognitive mechanisms remain unclear. In a simulated air traffic control conflict detection task with a 2 (Time Pressure) × 2 (Automation Reliability) within-subjects design, we integrated behavioral analysis with linear ballistic accumulator (LBA) modeling. Five competing models were constructed to identify how reliability affects cognitive processes under time pressure. Model comparison demonstrated that the drift rate model provided a decisively superior fit: low reliability under time pressure significantly suppressed evidence accumulation rather than altering decision thresholds. Individual-level analysis further revealed cognitive strategy heterogeneity—56% of participants were primarily drift-rate regulated, while 36% were threshold-regulated. Differentiated interface design recommendations are proposed for operators with distinct cognitive profiles.