To me, review scores are a double-edged sword. By having them, you have a number that is supposed to be an accurate representation of the overall quality of the game (gameplay, graphics, value, etc.) and gives readers a quick representation of how good they can expect the game to be. On the other hand, opinions between reviewers differ, and can result in very skewed scores that may not accurately represent the game. If one reviewer needs next-gen graphics in all of his or her games, but reviews a nice looking, but old-school 2D game, he or she might dock an extra point or two because the graphics aren't shiny enough, or vice-versa (a reviewer loves 2D graphics and adds a point or two extra because of it). In addition, if there's a reviewer that hates or loves a certain series, bias will be reflected in the score.
However, if you don't have scores, then reviews can seem abstract and it will be hard to quickly determine what's a good game, or even a good review site. A game could be almost perfect, but if a review nitpicks and writes a half-page of minor flaws, it would make the game seem worse than it actually is. If a game is completely horrible, but the reviewer praises the innovation and certain gameplay mechanics, then it makes the game seem better than it actually is. Without a baseline score to give a reader a general opinion about the review, you'll have a bunch of people who'll buy bad games (and may encourage more bad games) and a bunch of people who'll skip good games (and may impact future projects for that company) because of the way the reader interpreted the review. Of course, reviews may end up influencing nothing but informed gamers (as the contrast between scores and sales between Enter the Matrix and Beyond Good and Evil show). Even if only a small number of people are thrown off by the absence of scores, that's still more people than if there were scores in the first place.
I believe scores are completely subjective and aren't absolute on whether a game is good or bad, so I prefer having them rather than not having them. Seeing a game that recieves many good scores will pique my interest in it, and I see that a game I was interested in recieves bad scores, it won't cause me to immediately drop it. In either case, I'll read through a few reviews from trusted sites to see why a game got a good or bad score, and I'll make my decision based on that. If a game is heavily degraded for having (slightly) underwhelming graphics, (slightly) slow loading times, and a few, minor quirks, but everything else seems good, I'll think about it and see if I can deal with the bad to enjoy the good. If the review had no score, then I'd have to meticulously analyze each review to see if the reviewer actually liked or disliked the game, as everyone knows how hard it is to convey true meanings (like sarcasm) on the internet. I'd also miss out on all of the good games that I had previously known nothing about.
In short, while scores aren't an absolute definition of good or bad, they are needed to give a general overview of the reviewer's opinion.