Good questions!
Q1:
wouldn't people just use the bots of fake accounts to 'follow' accounts to game the system?
A1: People try anything so I'm sure they will try this as well! But as I've explained in the comments, the incremental value every follow from account X passes on, is decreased when X follows more accounts.
Say your absolute UA is 100 "points", and you follow 200 accounts with it, then 100/200=0.5 is passed on to every account you follow.
Now, suppose 10,000 bot accounts are created to all follow 1 account. Each and everyone of those 10,000 bot accounts receives close to zero themselves, so they pass along close to zero to that 1 "real account". Then, if that 1 "real account" only receives follows from those 10,000 bot accounts, wouldn't it be extremely obvious that it's an account swarm? I could even visually show those relationships.
"raw UA", and there's "Community / dApp / SMT / service" implementations" based on UA. Blacklisting based on certain behavior, such as you were accurately describing, is one of them.@cryptoctopus talks about blacklists and whitelists: a scenario such as this would be an extremely good candidate to get blacklisted for. There's
Q2:
users who have lots of followers already will always be ranked with high content, even if the content is not worthy.
It's not about the amount of followers, but about how much each of those followers contribute "UA-wise". 1 Follow from a high-UA account can easily contribute more than 1,000 low-UA follows from bots.
Also, "New rules, new games, new behavior": UA is dynamic, with eery follow broadcasted **all accounts' UA scores change! People can follow and unfollow as much as they like to, when they agree or disagree with another user's actions. It's a democratic system. Those users received their follows for what they did in the past. If they start to behave weird, they will most probably get unfollowed.
Q3:
Are there other platforms that have successfully use this technique?
Nope, it's created by Moi! and it took months to develop. My original articles were referenced at the top of this article.
"PageRank-based models" do exist for quite some time (since 1998 at least when Larry Page co-founded BackRub (later: Google) with Sergei Brin). But UA works differently nonetheless (also its technical & mathematical implementations differ, but both models are probability distributions).
I do like the idea. It has serious merit. For Q2, perhaps run an analysis to show the weight associated with accounts that have large numbers of followers. That will uncover if the 'UA-wise' factor will be a problem.