I heard about this before but completely forgot.
Getting honest answers about behaviour that is illegal or frowned-upon – such as taking drugs or visiting prostitutes – is notoriously difficult. But survey researchers have devised a neat way to get people comfortable with revealing their indiscretions.
Each time the researcher asks the respondent a question, the respondent throws dice before answering – crucially, the researcher cannot see what numbers come up. The rules of the game will be something like this: the respondent will always answer “yes” if they throw a six and “no” when a one comes up, but should tell the truth otherwise.
Because a “yes” doesn't necessarily mean that the respondent actually committed the undesirable behaviour, people seem to open up. The forced “yes” and “no” answers introduce some “noise” into the results, but overall this “randomised response technique” (RRT) gives better answers. For instance, RRT questions get much closer than conventional surveys to the actual incidence of drug use that is revealed by screening tests on hair samples.