I believe we should have the students specify the null and alternative hypothesis even before they start looking at the data. I think having them look at the data first might give them the wrong impression that you are creating your alternative hypothesis based on what you see in the sample rather than based on what theory/context would dictate.