A research method where participants are randomly assigned to use AI or not, to fairly measure the AI's actual impact.