This "peer reviewed algorithm" is basically PageRank, or close enough for the following critique to apply: None of these algorithms are models that infer trust from raw content, aka there is no special sauce for trust here, it's just an information theory problem in the abstract. So Garbage In, Garbage Out. These algorithms are only good when you have a single, consistent signal of trust that you can trust. So it will have the same issues that Google has with SEO spam, and there's no graph-traversing algorithm to solve that, as it is an out-of-band issue compared to what these algorithms are designed to take as input. Take outgoing links to be your signal of trust, and you'll create a linkspam problem. So you need an anti-spam model that sits in front of this algorithm to detect violations of webmaster guidelines, aka gaming the system.

So in other words, all you need to do to game the system is generate a bunch of accounts that publicly link to you, follow you, or whatever this systems' input of 'User A trusts User B' is. And not get caught by whatever anti-spam or anti-bot model they're using.