Fig. 1

Experimental designs. Note: This figure presents example tracks for the three separate research designs used in the experiment. The abbreviations represent the assigned information track. XO corresponds to x-ray only, CH corresponds to x-ray with clinical history information, AI represents x-ray with AI assistance, and AI + CH represents x-ray with AI assistance and clinical history information. In the first design, each radiologist is randomly assigned to one of 24 possible sequences of four information conditions. Within each condition, they evaluate 15 unique patient cases, and no radiologist sees the same patient more than once. In the second design, each radiologist evaluates a total of 60 patient cases across four distinct information environments. The experiment is divided into four sessions, with each session separated by a washout period of at least two weeks to minimize recall of previous assessments or AI assistance. In each session, radiologists assess 15 patients under each of the four environments, encountering each patient only once per session. In the third design, each radiologist reviews 50 patient cases, first without AI support and then with it. Clinical history is included for half of the cases, with its presence randomized to either the initial or latter portion of the session. The sets of patients evaluated with and without clinical history are distinct.