In an experiment on social feedback, participants cannot tell whether feedback is real or computer-generated. The experimenter delivering the feedback is blind, but a second experimenter who knows the condition later conducts interviews that become the main outcome measure. Which evaluation is best?