I think it depends on the study matter and double blind is not always necessary. Yet there are certain cases where double blind would be preferable. It is used to prevent research outcomes from being influenced by the placebo effect or observer bias. If a researcher knows that certain patients are getting the actual drug (lets say for Alzheimers), they may subconsciously emphasize certain questions or invest more time to help the patient score improved outcomes. Also there is research that shows that when a researcher invests more faith in the patient, the patient can somehow detect this and this feeling boosts their outcomes. Thus to prevent any subconscious (or conscious) bias, it is better to use a computer or double blind method when there are high stakes.