Testing if people like one thing than another

I want to test given a certain scenario, if people will like one thing than another.

Say there are 20 participants, 7 of them will chose A, 10 will choose B, and the rest 3 will choose I don't know (give up).

I need to test, if people choosing B is significantly more than people choosing A.
You have binary choice from a relative small sample. The best you can do is a chi-square test goodness of fit. It doesn't suggest a causal relationship either.