I disagree. First, I would always go for two when down two It's awfully presumptuous to act like it's a certainty your team will score again. Of course, failing to go for two will bite you in the ass only if (a) you don't score again and (b) the opponent doesn't score again, but you never know. Down 14-12 in the 1Q? Extremely unlikely that it would matter, but I'd still go for two. Down 14-12 in the 3Q? More likely to matter, and again I'd go for two.
As for your second paragraph, I've long been a proponent of going for two when down by 8 late in the game. If you fail, you still get to try to tie it on the next TD. If you make it, you just kick the PAT next time, and you win. The odds make it the expected outcome superior because you are accepting the chance of a two-point loss in exchange for a greater likelihood of a one-point win.