This question was raised by Dave Clarke here.
When a textbook author approaches a topic in a novel way or presents a particularly interesting example, I believe that a teacher who creates lecture notes using this novel approach or interesting example would be doing the right thing to cite the originator of the approach or the example. Similarly, whenever I copy a clever (and clearly unique) problem from a textbook and give it as a question in an exam or an assignment, I try my best to indicate (in the exam or assignment paper itself) the source of the original problem.
Does anyone know of any written document indicating whether or not it is considered unethical to copy a published problem and put it in an exam or an assignment without citing it?
I think it's considered more of a courtesy rather than a requirement to credit someone who has developed a problem, provided that there is no new technical content introduced in the problem. On the other hand, however, if one is to use a problem in a problem set or examination completely unchanged, then some citation of the original source is certainly recommendable, as otherwise one is guilty of a copyright violation.