The hypothesis testing approach to statistical inference, which is what most of the scientific community uses, explicitly demands researchers try to prove that, say, AGW has no major impact, before reporting on their failure to do so.

The hypothesis of a major impact is suggested because negligible impact cannot be, in the face of data. The popular Neyman-Pearson test doesn't even account for failing to detect an effect - it concerns itself strictly with not seeing ghosts.