Answer
I would use a random sample.
Work Step by Step
The systematic method may create biased results. For example, if I asked every 10th person who walked into the cafeteria, the results may be biased. Perhaps, people who listen to a certain type of music are less likely to eat lunch in the cafeteria. A stratified method would work, but it may be difficult. After all, going into each classroom and randomly selecting two students would a) take a long time and b) be disruptive.
When writing a survey question, there are a variety of important considerations. My main advice would be to make sure that the question does not in itself suggest a certain answer.
