You don't even have to force people to behave one way or another. You can skirt around the ethical issues by giving Sample A an unlimited supply of free masks,

You don't even have to force people to behave one way or another. You can skirt around the ethical issues by giving Sample A an unlimited supply of free masks, and keeping Sample B as the control sample. Even with free masks, not everyone in A will wear a mask. And some in B might still bring their own masks. But A will have higher mask use on average, and with large randomly selected populations you can tease out the impact using IV regression[1]. [1] https://en.wikipedia.org/wiki/Instrumental_variables_estimation https://en.wikipedia.org/wiki/Instrumental_variables_estimat...