The Gorilla in Your Control Room: Inattentional Blindness in Monitoring Work

Attention is a filter, not a camera

The intuition this demolishes is that seeing is passive recording. It is not. Attention actively selects, and what falls outside the selection is not dimly registered — it is frequently not registered at all.

Every monitoring role in your organisation rests on the opposite assumption: that a person watching a screen will notice a significant anomaly because it is significant. Simons and Chabris demonstrated that significance is not what determines noticing. Task-relevance is.

Alarm fatigue is the industrial version

In plants, hospitals and network operations centres, the failure mode is well documented. When alerts are frequent and mostly non-critical, operators do not merely learn to dismiss them — attention reallocates away from the channel entirely, and the genuine event arrives on a channel nobody is watching.

The standard response is to make the critical alert louder or redder, which raises the baseline and accelerates the same adaptation. The effective response is to cut alert volume so drastically that attention to the channel is worth sustaining.

Designing monitoring people can actually do

Give the monitoring task a structure that keeps the relevant features task-relevant: active sampling and periodic verification rather than passive watching, since attention sustained on nothing degrades within twenty minutes.

Rotate monitoring duty. Use automation to filter rather than to alert — reducing what reaches a person is more effective than flagging harder. And in any incident review, resist "the operator should have seen it." Half of Simons and Chabris's observers should have seen a gorilla. Design for the attention people actually have.