First, again: not all polls. There is such a thing as a strictly scientific approach. A proper design of experiments will enable one to construct a poll that approaches true objectivity.

However, I grant the big "IF". Still, it's been done. I know. My polls are never used to sway opinions.

Second, there has to be an assumption of honesty. And guess what? For the most part it bears out, especially the more anonymous the poll is. People tend to be more honest in well-constructed polls than in almost any other venue. Margin of error accomodates dishonesty.

I'm thinking your exposure to polls has been somewhat limited. I'm just glad I'm not the most cynical person in the discussion.