Selective Prediction Reduces the Negative Effects of Automation Bias Overall but Increases False Negatives
The episode discusses a paper titled "Selective Prediction Reduces the Negative Effects of Automation Bias Overall but Increases False Negatives." The study, involving 259 clinicians, found that hiding incorrect AI predictions improved overall accuracy but increased false negatives—undertreating patients. The hosts conclude that selective prediction should not be a default safety mechanism and requires careful context-specific evaluation.