For example, do you more or less likely to upvote the currently high score answer? Do you ever upvote a question or answer because you think "it does not deserve negative score." although you don't particularly like it either.
This really is a psychology question. One has ones ideals and principles and then there is reality.
Ideally, I try to ignore how many votes a question/answer already has. I believe that (ideally) voting should always be based on the content and not other external things. I do not factor in whether or not other people have already "liked" the post, but I just ask myself if I find the post worth my vote.
If the question is interesting and isn't just the statement of a problem, then I will upvote. I might also upvote if I think a question is well worded and shows the context. So, I might see a basic calculus question where the OP is clearly expressing his/her confusion. I never vote based on who is asking/answering. From time to time one might come across a user who in general asks "bad" questions and it is tempting to downvote everything this user posts, but I refrain from this.
I often also upvote questions that I am answering. If I think a question is worth answering, I think it deserves and upvote.
Sometimes I see a "hot question" in the side bar and from there I am directed to the question. I will then sometimes upvote the question and the best answers.
I also often upvote competing answers.
I mentally assign scores to answers, and my probability of upvoting the answer depends on how much its actual score exceeds the score I have assigned. I am very unlikely to upvote an answer that I feel has more than enough points already, and I am more likely to upvote an answer that I feel is under-appreciated. I'm not sure this practice is defensible, and I don't mean to defend it here, or even to recommend it.
I do not downvote questions just because i think their score is too high; I only downvote answers that contain significant errors.
My voting habits for questions are very different.
Yes. Three examples:
Often, questions are closed for being poorly posed (because, for example, the OP didn't understand the site), but they are improved and re-opened. When this happens, I believe that the barrage of downvotes is unhelpful - they were intended for an old version of the question, not for the new, nicely-worded version. So I upvote. My upvote is not for the brilliant question, but rather for "balance" - ideally the downvoters would remove their downvotes (so no "balance" would be needed), but this rarely, if ever, happens.
If a question receives an upvoted answer then the question is removed from the "unanswered" queue. Therefore, if I see an interesting question with a half-hearted/mediocre answer (subjective!) which doesn't actually answer the question (perhaps it addresses a boring special case) and the answer is on +1 then I will downvote. This puts the question back in the "unanswered" queue, where hopefully it will get some more attention.
Suppose an excellent post comes late to the party, and although excellent it is never going to catch up with the other two equally excellent answers, each of which are on +8. So I upvote the new answer and do not upvote the other two answers, even though they are deserving of my upvote. Again, this is for the sake of "balance".
My up/down votes are independent of the current score of the post. The content of the post is what ultimately influences my choice to either vote up or vote down. I'm much more likely to up vote a late post that thoroughly answers the question as opposed to an early post that just barely scratches the surface.
I vote up if the question or answer gives me something of value. Since I don't know why others did vote up or down that is uninteresting for my voting.
Some answers are great, and most readers of such an answer would be inclined to vote it up. Moreover, such answers correlate with good questions, which are highly up-voted and so attract more attention. Thus one might expect to see answers with hundreds of up-votes. I suggest that the rarity of such a score is because voters take the existing score distribution into account. Perhaps the thought is "This answer is in the top 2%, but not in the top 0.1% as its present score implies. I don't need to add my increment to its overvaluation". Conversely one might think that a good answer deserves a higher rating than it has presently.
These are merely my own thoughts; I am happy that others think otherwise. For good reason, there is no guide set down on what to consider before voting, not least because any such guide would (in my view) be absurd, unimposable, and much disputed.