I think this applies to almost all online rating systems, in particular ones where it's just pick a number out of five or ten. Most will be either very high (perfect, or perfect with a slight problem), or the lowest (completely shit). Barely anybody will mark something as 2/5 because it was utterly terrible, but this one feature was good. They'll remember the bad parts. No, no data at the moment and it's just a hunch, but one I hope to be able to measure soon.
I'm hoping to apply this to webhosting, rather than simply rate out of ten and leave a 2 sentence review, it'll be much more indepth, with multiple options (support, uptime, features etc). I'm betting that the more effort that you go through to leave a thorough rating will be balanced by the fact you won't have to rate that often (at most maybe once a year once you find a good host).
Oh well, still in the idea stage but I think people should reconsider just a simple rating out of five for anything non-trivial (like rating videos). Anything important (major purchase decision) should be more detailed and have more information for the prospective user.
Videos are a bit different, because people review as they are watching. However with something like hosting, you need to give considerable thought to how you get people to report positive or average experiences equally as often as negative ones.
That's just my initial reaction.
It's definitely something to think about and it's a pretty difficult balance I'll have to achieve but I'm confident I can figure something out.
YouTube noticed a similar issue. Getting users to rate on a star rating just isn't effective. You may have more luck with a simple 'yes/no' option. Would you make this again? And offer the opportunity for comments.
I find the comments in the reviews on epicurious to be very influential in how I follow the recipe, and if I cook it at all.