Complete bounded explanation
What this power means.
Capability to develop or express visual search in a declared context without inheriting broader claims.
What the current evidence supports
Visual search is trainable for defined targets and contexts, yet low-prevalence targets remain easy to miss and broad transfer is not guaranteed.
How to observe or measure it without overclaiming
Predeclare target class, prevalence, sensitivity, false alarms, response time and unfamiliar-scene transfer; keep speed–accuracy trade-offs visible.
Human anomaly scanner
MetricSensitivity, miss and false-alarm rates on rare unfamiliar targets
BoundaryFast finding of frequent trained targets does not establish expert anomaly detection.
Negative and limiting findings
- Rare targets were missed at disturbingly high rates.
- Contextual learning can improve a repeated scene without improving a new one.
Claims this evidence cannot support
- Photographic scanning
- Never miss an anomaly
- One game trains every search job
- Reaction time proves safer inspection