Why not, if you would get reputation points for it? The idea is to review the papers which you want to read anyways, or perhaps you have already read them and you want to express your feelings about these papers, hopefully in a way useful to other viewers of the paper. Nobody tells you to do it, so you do it by heart, this surely would increase reviews quality.
Well, maybe this could work, but probably it would have to replicate the "coercive" component of the current system. In other words, by reviewing other people's papers you somehow improve the chances of having your own papers reviewed. And if somehow that system manages to also provide an incentive to do the job thoroughly, then that would be a huge improvement on top of what we have now.
Well, the coercive component is the unified reputation that you gain in this site, just like the reputation that you gain in SO. Obviously, the higher reputation you have, the more recognizable you are and more people read about what you do.
