Yeah, that’s the same problem in fact - data spam or ratings spam.
I was thinking about the same topic too.
My conclusion was that usable decentralized systems can’t exist.
Then I looked at Yggdrasil, Tor, I2P and found that they are actually fine.
Maybe bot owners are too lazy to target such projects, or they found some solutions.
Worth investigating I think.
Idea is to make basic features of network powerful enough to allow making filtering system and then wait until users find out how to actually make it.
I know that many people won’t like it, but I don’t see other possibilities.
Making flawed system which with high chances will instantly fail is not the route I agree with.
I remembered something.
Yggdrasil allows to mine identity.
And then anyone can check how “strong” it is.
Maybe identity PoW instead of rating PoW can be useful.
I could see ratings working if user is requested the first time to eg authorize enough funds for ~100 ratings (and informed that is say .01 USD equivalent) and then not bothered again until all the funds have been used up. But if one has to authorize for each rating, that’s too high a bar.
Honestly I see that in order to rate you have to at least join as a member, then maybe a system where you have to be active before you can rate. And maybe the only rating that matters is from those who are more active then a casual person.
I think a fairer system would be that the user can “subscribe” to various other users and then only their ratings matter. Maybe a averaging or Mean and fall back to global if uncertain result from the subscribed to members ratings/votes.
There could even be suggested groups of members to “subscribe” to.
This means that spamming only affects those without subscriptions or visitors. Maybe the forum owners can choose a group to use for those without subscriptions, basically a default group of raters.
This also is a way that moderation if desired by the user can be implemented too. So rather than a set group of moderators like here in this forum, the user selects who to subscribe to for moderation effects. While some will want no moderation, aka wild west, others will want at least the minimum of no spammers/vile illegal stuff
Just so you know I was meaning the record store cost not a UI. But I suppose that is possible too
What if ratings were nanopayments. So turned into rewards where we can see from the pub key how many (not just how much) rewards were paid. Could solve the bot issue, and some folk might want to inflate their rating by spending cash on themselves but that might become a zero sum game for them?
The concept I came up with for my social media app idea was that people would submit tags with their post; then people who see the post would pay to rate each tag as well as a general rating of the post. The cost of each rating goes up exponentially as the sum of both likes and dislikes increase - this hammers paid bots. These payments go to the poster minus whatever the data cost is. So all of the money flows around the network.
I see this as a meritocratic system.
The users client would also filter and sort based on these tags and general rating.
The differential between the likes and dislikes can also allow filtering/sorting based on controversy - but all of this would be at the users discretion.
I guess a TL;DR summary would be: it’s very complex, and there are no easy answers, much as we would want there to be. But it’s well worth a listen, because it gets down into the weeds of the trade-offs of varying approaches, how the nature of threats change depending on where and at what layer lines are drawn.
heh… now we are getting into “web of trust” models… something that was hyped in certain circles and I was deeply exposed to for a couple of years before deciding it was computationally more trouble than its worth.
Basically the notion that every ever vote/rating would be weighted by how far away they are from you in a graph of your trusted sources.