"No one seems to question the dataset".<p>Actually, this has already been pointed out in papers written on the subject. If you want to learn about the Netflix prize, read the papers from the teams that solved it, and don't waste your time on self-important bloggers who want attention.
YouTube recently admitted that they have a similar problem with their rating system (<a href="http://youtube-global.blogspot.com/2009/09/five-stars-dominate-ratings.html" rel="nofollow">http://youtube-global.blogspot.com/2009/09/five-stars-domina...</a>)<p>The raw data simply doesn't show a decent distribution of ratings. People are far more likely to simply give something 1 star or 5 stars.<p>I would say that it's far more meaningful to use a simple thumbs up/thumbs down system...