RE: Is there a more useless metric than reputation?
You are viewing a single comment's thread:
I've done some experimentation in implementing web of trusts before, and while I've never managed to get anyone else to implement any such system in a live platform, the concepts remain more or less the same.
The biggest problem for such systems is that they're extremely computationally intensive as difficulty scales exponentially with network size. Especially in the case of trust propagation.
Trust propagation calculation would provide trust levels that mimic how in person communities trust each-other. It'd be vulnerable to the same attacks as such too, but should be immune to many common digital attacks such as someone creating a thousand bots that then upvote each-other to infinity.
As a less computationally intensive alternative, Hive could implement a compromise of calculating per-community reputations. Hive consists of many communities, and it would make sense for people to have their own reputation score within every community in addition to their global reputation score. This would partially solve the issue of reputation not being a good indicator of behaviour in a given area.
Would it be simplified ( :D) if there were many webs calculating only segment of the data and then being collated and factored together, without needing to calculate every factor individually?
BTW, I am definitely not the person who should have influence in the way to do this :D
Back in the day, the User Authority score was doing something complicated (very flawed imo) but had a decent idea behind it.
I do think that communities should start adding in their own metrics that can be leveraged by the entire community also. I think that at the very least, it is a layer of gamification that gives people targets within communities to aim for and could even be used to raise people into positions like moderators, without relying on "who you know" as much.
Thanks for dropping in - it is a really interesting area, despite me not having a good grasp on the technicals.
There are many tricks you can do in order to reduce how much time is spent on calculation, it will still scale exponentially though as fundamentally you're calculating a unique user score for every user per user. A network of 10 users will have 100 scores, a network of 100 users will have 10000 scores.
What most programmers do in the case they've got something that grows exponentially is that they'll set a hard cap on how many items can be processed together. For example, Discord limits users to a maximum of 100 Guilds because one of their algorithms scales exponentially in computation usage. In Discord's case however it's likely down to bad programming, their entire platform is a disaster from privacy to moderation management to developer competence.
I've got a lot of projects in mind for Hive. Though I have no idea how many, if any, I'll ultimately implement. The blockchain does lack some features that would make my ideas easier to implement, such as on-chain smart contracts, but we'll see where things go.
I think that all experiments are welcome on Hive and at the second layer, there might be some very useful implementations we can all learn from across multiple points. At the very least, this is far more fun than posting pics of meals on FB :)
I'm mostly here for the technical aspects of Hive, I'd like to see it become popular enough to weaken the tech giants like Google and Facebook.
There's a bit too much focus on rewards for my liking, it's to a degree that'll potentially scare off a lot of users. I made a post a short time ago regarding some major things Hive needs before it can seriously grow, I probably should've mentioned the over emphasis on rewards somewhere.
I think that the rewards are a novelty that takes focus as it gives everyone the ability to earn something. This will likely change in the future as there is more spread as well as more people mainly looking to consume.
!ENGAGE 20
ENGAGEtokens.