Never rely on human vigilance for low-frequency monitoring…
When sustained attention must monitor for low-frequency errors over extended periods (>30 minutes), delegate detection to automated systems rather than relying on human vigilance, because attentional resources degrade predictably regardless of motivation.
When This Fires
- When designing any monitoring or surveillance task that requires sustained attention
- When humans are assigned to watch for rare events over extended periods
- When monitoring failures occur and the root cause is "nobody noticed" — the vigilance decrement is likely
- When deciding between automated alerts and manual monitoring for system health
Common Failure Mode
Assigning a motivated person to the monitoring task: "Sarah is diligent, she'll catch any anomalies." Sarah's diligence doesn't override her neurobiology. After 30 minutes of watching a dashboard for rare anomalies, her detection rate has dropped below 60% regardless of how much she cares. The fix isn't a more motivated person — it's an automated alert that fires when the anomaly appears, regardless of when it appears.
The Protocol
(1) For any monitoring task, assess: how frequent are the events to detect? How long is the monitoring period? (2) If events are infrequent (fewer than 1 per 10 minutes) AND monitoring exceeds 30 minutes → automate detection. Do not rely on human vigilance. (3) Design automated detection: define the signal, set the threshold, configure the alert. The alert should be unmissable (When triggers fail to fire, increase signal strength…) — it replaces the vigilance the human can't provide. (4) Reserve human attention for alert response, not signal detection. When the automated system alerts, a human evaluates the alert and decides on action. This is contextual judgment (Automate pattern-based error detection before manual…) — the human's strength. (5) If automation is impossible → implement mandatory rotation: no individual monitors for more than 20-30 minutes before being replaced.