Industrial inspection — detecting defective products on a production line — is a signal detection task where the consequences of misses (defective products reaching consumers) and false alarms (discarding good products) have direct economic costs. SDT provides the framework for understanding inspector performance and designing optimal inspection systems.
Prevalence Effects
In quality control, defect rates are typically very low (often below 1%). SDT predicts that low signal prevalence leads to conservative criterion placement — inspectors become reluctant to call items defective when defects are rare. This "prevalence effect" has been confirmed experimentally and has parallels in medical screening (e.g., rare cancers being missed in mammography).
When defect rate = 0.01 and costs are equal:
β* = 99 (extremely conservative)
Vigilance and Automation
Extended inspection shifts produce vigilance decrements — declining sensitivity (d′) and increasingly conservative criteria over time. SDT analysis distinguishes these two effects, which have different remedies: declining sensitivity may require rest breaks, while criterion shifts may require feedback or adjusted payoffs. Automated inspection systems can be evaluated using the same SDT framework, with human-automation interaction modeled as a joint detection system with combined ROC performance.