Forgive me, as I'm probably misunderstanding this, but surely there's still a bias in the randomly selected poll, because then people who think that their "subject" is inherently of worth will vote, whereas those who are ambivalent won't?
I think I'm getting confused by your assertion that "people choose to vote", where it seems you're implying that the sample group is chosen by the vote-taker, because they might be biased in favour of pre-conceived ideas?
Of course, there is always bias in polls. Even ones run by companies that do it for a living and have a vested interest in getting it right. That's why you see a +-X% polling error next to any poll. The thing is that you follow best practices to reduce that number. Also every polling agency has their own internal formulas that take into account how many people are actually going to vote on election day and how to get the numbers they asked into a sample that accurately reflects the electorate.
Good polls aren't "We picked 500 random people and this is what they said....." they try to mirror as closely as possible the people they are trying to represent, which something on a website like this just can't do.
Done well, polls are amazing when it comes to predicting things, done poorly they are no better then a magic 8-ball.
There was a reason that 538.com was able to predict how every single state in the last presidential election would go, he even predicted correctly which ones were close. ((Florida and Virginia being basically down to the last vote counts.)) It's because he understands polls, and what they say and can tell you.