I think it is easy to be cynical and easier still to say, "Just fix it, nerd," but trying to detect collusion between two parties when you do not control their

I think it is easy to be cynical and easier still to say, "Just fix it, nerd," but trying to detect collusion between two parties when you do not control their communications is quite difficult, especially when their associations will be only fleeting. For example, you could disallow changes in star ratings, but that would also prevent people from rating something only to change their minds when it broke three months later and you've lost useful information that way. You could make sure all communication goes through Amazon, but that only counts for email. The seller has your physical address and therefore can offer you gift cards to change your mind, via a postal offer. You could hide the names of reviewers, but then you risk losing the trust metagame where there are reviewers people trust. You could allow only the verified purchaser of a product to comment, and only once, and that makes "review bombing" of competitors expensive but not impossible. I imagine you would need a good outline of all types of different "fakes" (I am sure multiple sorts exist) and scams, then find the nastiest game theorists available and discuss what could and could not be done. The more I think about this, the more Amazon Marketplace morphs into Amazon Escrow.