Miscalibration

Biological Parallel

Cheetahs exhibit systematic miscalibration in hunting success estimates—they persist in chases where success probability drops below energy return thresholds, suggesting internal confidence models don't match reality. This miscalibration persists because occasionally successful 'long-shot' hunts provide enough reward to reinforce overconfident pursuit behavior. Perfect calibration would require constant Bayesian updating with complete information, but ecological conditions change faster than learning algorithms can track, making persistent miscalibration nearly inevitable.