This depends to some degree on the individual case, but I think the problem with these questions is, uneducated readers could take the wrong assumptions for granted, especially when there is no top answer explaining the misconception, or the wrong assumptions are not obvious. A downvote is probably the easiest way of pointing out there is something wrong with the question without having to go into a long-winded discussion with the OP.
I have often stumbled upon posts, where if the asker had taken a little bit more care, he/she could have asked essentially the same question without making the wrong assumptions, or by stating clearly he knows he is making an assumption which might be wrong. For example, instead of asking something like
"I am trying to use OOP to solve problem X, and since the OOP way always must involve inheritance, I did it using this class hierarchy, which led me to problem Y. Is OOP broken?"
a better question could be
"I am trying to use OOP to solve problem X, and I was assuming inheritance might be a good idea to use, I did it using this class hierarchy, which led me to problem Y. Is there a better approach?"
So I think the appropriate action is
- when you see such a question, and the wrong assumption is not a minor issue, downvote, leave a comment and ask the OP to edit his question
- if you have have enough rep, and you think the question is worth it, try to fix it by editing the question by yourself
- come back later to the question and remove the downvote after it was improved
(However, some questions are so badly written or so full of wrong assumptions they cannot be saved easily, so for this cases I recommend downvoting and probably close-voting as "unclear").
There is one issue with this, people often tend to take downvotes way too personal. So when leaving a comment after a downvote, one should be kind, make sure the OP gets to know how his/her question might be saved and encourage him to edit and improve the question.