Well you could take two approaches.
If you have a databank you can poll you can implement a constant time algorithm. Essentially if you database you are polling is storing hashes of the key, in this case the question, then the value returned should be the answer. If we set our databank to be an instance of Google, then if we poll Google for the answer, it should work.
Or, you can do an exhaustive search. There are only 4 possible answers. 3 if you remove the answer you have already determined to be incorrect. So brute force should complete in a meaningful time. In this case you can implement a linear time algorithm bound by the number of possible answers that would attempt to answer the question until it finds the right one.