This seems like an observation about either pointers or Meta-questions. If I pick an answer to {an unspecified thing}, what is the probability of being right?. Choose A/B/C/D.
First, I have to assume the {unspecified thing} is your question itself. Then model a clone of your question, then based on the answers to the clone model, answer your real question. Then I have to make some assumptions about the biases in my random selector (i.e. it has none and picks each option on average 25% of the time). Then I have to make some assumptions about whether one of the answers is right and three are wrong, or if more than one can be right.
If I assume one is right and three are wrong, then A and D are ruled out, so the right answer must be B or C. C can't be right, so it must be B, so my chance of picking "B" and getting the right answer is 25%. So I answer your real question either A or D.
If I assume that A or D could be right, then I assume B and C are wrong and get a 50% chance of picking correctly, so I answer your real question B.
I guess I could carry on and say "if no answer is right and I always pick one of the four, then I have no chance of answering correctly and answer your meta-question with C" and then see if I can sum up all the potential assumption answers and see if it resolves to a most likely sum-over-histories style answer.
But that sounds like work.
Selecting an answer at random, in my dialect of English, can be said to be choosing. In my dialect of English "to choose" does not necessarily imply the exercise of free will and decision.
Thus, "to choose at random" makes sense, and might be regarded as "to select at random" without any implication of a deliberate decision (except the decision to follow the random "choice").
Does that answer your question?
But I've expanded my reply so it's less relevant now. [edit: oops, or replied again].
I cannot understand the point of hiding deadened questions while still keeping the comments?
It makes sense to me. If the users flag & kill a link submitted by spammers, the spammers get no value from it because, by default, a spider won't be able to follow the HN page to the URL the spammers wanted to promote.
But keeping the comments helps users at HN get a feel for how the article was initially received.
It's all about removing the visibility. If the story is hidden, so are the comments, and so there's no reason to [dead] all of the comments too. The only way to see those comments is if you have showdead on anyway...