We developed and tested a classroom experiment on 224 medically trained students. Short texts with ambivalent patient descriptions were presented, and participants had to state confidence in one of two diagnoses.
We found confirmation bias in selecting follow-up questions, and increased confidence when not exploring. We did not find confirmation bias to be associated with confidence, and no anchoring effect of symptom order.
Our aim was to test the traditional decision-making biases in more ecologically valid settings. I will claim it as a mixed success. Conceptual #replication in is difficult!