You are viewing a single comment's thread from:

RE: UserAuthority (UA): explanations, applications and implications

in #utopian-io7 years ago

These were two of my favorite articles I've come across. I was writing an article and was struggling with whether or not this was possible when I saw your first post telling me it was! I think we could extend this by adding a few more metrics, if people have muted you it would be a negative score, if you upvote spam its bad, if you downvote spam it's good. It would also be able to identify what posts are spam and which are not. Anyway, thanks for the article! I linked to it in my post and used your diagram, if that's not ok let me know and I can take it down. I just loved it! https://steemit.com/steem/@littlejoeward/fixing-the-reputation-score-and-eliminating-spam-a-follow-up-to-user-authority

I would love to work with you more on this idea, though my coding skills aren't very good (beginner) I might be able to help with testing or I'm pretty good with formulas on google sheets to create some simple tests of these ideas. Anyway, let me know if you need anything!

Sort:  

Thx, and you're welcome! Read your own article as well. Could you maybe mention my account name in there? (@scipio)

PS:

  • calculating UA in a blockchain decentralized way is in fact very hard to implement, but I found a solution for that already, screen this comment thread for my reply to @personz : the solution is in my response.
  • there are many extensions possible, using UA at its core to construct all sorts of new metrics.
  • incorporating all upvotes/downvotes, mutes, content-analysis within UA itself, would make it impossible to be implemented in real-time for blockchain use.
    So, nice ideas, but UA needs to stay elegant and small for computational reasons! What you want is all possible via extending UA with other metrics.

PS: upvoted your post as well

No problem... and done.

I've read every comment on both of your posts! Very interesting discussions! I was glad you came up with a solution because I was afraid that it was impractical to implement. A couple ideas: (sorry if they are dumb ones haha)

I realize that I am probably missing something but it seems to me that if you are already looking at all followers it would only increase the computational power by ~4x by looking at mutes/ upvotes and downvotes on accepted/spam posts. Wouldn't you use the same basic process?

Could we just calculate it once per week or per day or something because it isn't really necessary to have it be real time because it shouldn't change much day to day.

Or maybe it could be something like the current reputation system (I think) where it simply is added to/taken away when an event happens. (someone follows/mutes you, your score goes up or down etc) It might not work quite as well but I think it would be much simpler to implement/compute.

The real innovation is my implementation for blockchain decentralization: how to embed UA steem-wide. The rest is just 1960s math, few seem to understand (as long as it gets implemented, I don't care).

Incorporating upvotes/downvotes inside UA is not feasible computationally. Because on average every account hold thousands of upvotes/downvotes.

Most importantly to understand on why UA is completely different to SP or Rep: it's backed by a probability distribution where the sum of all UA = 1 (100%).

Coin Marketplace

STEEM 0.19
TRX 0.13
JST 0.030
BTC 60543.19
ETH 3336.30
USDT 1.00
SBD 2.48