Could @steem-ua + @freezepeach form the perfect combo?

in steem-ua •  6 years ago  (edited)

My first impressions regarding @steem-ua are rather mixed. On the one hand, the existing reputation system is horribly broken. Utterly destroyed by bid bots, orca/whale self-votes, and other false-curation leverage tools. So yes, we really need a better reputation system. On the other hand, though, UA also comes with its own false curation, potentially making the underlying problem, stuff getting upvoted without an actual user looking at it, increasing the legacy reputation of the creator unduly, even bigger.

So why exactly is a broken reputation system a problem? Well, the reputation system mainly serves a purpose in allowing for self-regulation on the platform. If I rent a shitload of SP, post crap-posts twenty times a day using some silly script and then upvote each of my own posts, I'm abusing the system and someone with a higher rep than me could (and should) attenuate my growth by flagging my posts. There are two parts to getting flagged. First, obviously, there is the downvote part of the flag. This does some direct monetary damage in reducing payout. Secondly, though there is damage to the reputation. Get flagged enough and your rep will go down so low that you become invisible. Abusers becoming invisible is an important part of the rep system. A part, though, that gets broken if we see how false-curation votes influence the ability to get flagged. Today the rep system is more of a reflection of the use of false curation than it is of actual reputation amongst steemit users. Bid bots and self-upvotes done with rented SP will boost your rep and in doing so will make you untouchable. A flag by someone who doesn't use bid bots or high powered self-upvotes is unlikely to have any effect as the false curation user will in most cases have the higher rep. That is, in a sense the legacy reputation system is inverted. The abuser who got by his rep through the use of bid bots and self-upvotes can (and will) retaliate flags by #nobidbot and #purecuration steemit users and because (s)he has a higher rep these retaliations will cut into the rep of the upstanding steemians. Even is the upstanding steemians have a higher rep, (s)he will know that the abuser will likely soon buy him/her-self enough rep to retaliate.

Ideally a new hardfork would :

  • Use a witness agreed list of false curation bots who's upvotes don't contribute to reputation.
  • Make self-upvotes and upvotes by related (proxy/recovery links) accounts not count towards reputation.
  • Do a full history reputation rollback on the entire blockchain

That way our abuser would be stuck close to 25 and upstanding steemians coold flag the abuser furhter down.

So given this, is it enough to have a better (out of band) reputation system? Well, I would argue that if this reputation system is used for yet more false curation, then no. There is I think another possibility though where UA could attenuate and possibly cure one of the worst consequences of the broken reputation system.

To get to my proposal, we need to look at a brilliant initiative by @r0nd0n, the @freezepeach service. The @freezepeach service allows those who are wrongly flagged to ask the service for a matching upvote to cancel or attenuate the damage by the flag. While @freezepeach isn't a highly powered account, even at its current strength it serves an important purpose. A service that serves only part of the flag/rep problem, but one that solves an important part.

Now let's focus on flagging and flag wars and see how UA could be changed to help there in a way that would reduce the false curation the UA service does itself.

My proposal is the following:

  • Make a semi-automated UA based @freezepeach bot. If a low UA account flags a high UA account, the flag gets queued for review.
  • Have @r0nd0n and his people moderate the flag queue. If they mark the flag, the UA service tries to cancel or attenuate the flag.
  • Give @freezepeach the voting strength of the UA service to use when needed.
  • Replace the algorithmic curation with an algorithmic vote follow. That is, don't upvote posts of delegators based on the UA of the delegator, upvote posts upvoted by and downvote posts downvoted by the delegators based on the delegator's UA score instead.

I think making these changes would both take away the problem with the use of algorithmic curation and be a major step in attenuating the problems caused by the legacy reputation system.

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  

I think one of the biggest issues I have with the UA score is how heavily it weighs just being a witness, it is not a real score of engagement, nor of authority, a few on the UA score system in the 7 + range should be in the minus range, but that is just my opinion.

There are two things that I want to bring up: (1) other stuff with the current STEEM reputation system, and (2) my own personal objections to the UA reputation system.


Another problem with the current reputation system is that it reflects votes coming from everywhere within the STEEM blockchain, so that number we see next to our usernames on Steemit will not accurately reflect our true reputation within a specific site or dApp, unless we are only using one such site/dApp. A reputation that is built entirely upon either only posting memes on dMania or only answering questions on Musing.io isn't exactly going to be useful in assessing a user's blogging reputation on Steemit. The more sites/dApps someone uses, the more difficult it becomes to assess.

Take a simpler case like my own: my current STEEM reputation level is 55 (and closing in on 56). But unless I publicly state it, most people wouldn't know that I basically have the equivalent of a 53 from over a year of blogging on Steemit, and another 53 from less than a month on musing.io. The 53 from Steemit isn't fully legit either because I did briefly experiment with bidbots and got a couple of significant-sized upvotes out of them, plus I'm enrolled in a couple of "long-term outlook" upvoting programs. So the reality is that I'm better at answering questions than I am at blogging, but no one would be able to figure that out by just dropping onto my profile page on Steemit. Some other site/dApp developers recognized this potential issue and wisely chose not to include STEEM reputation in users' profile pages on their sites for this very reason.


Now as for the UA reputation system, I first need to go back and expand on a comment I made in one of your previous posts about the price of STEEM going low -- if you'll recall, I mentioned a psychopath whale without naming them for obvious reasons. There are actually 5 users who I refer to as the "Psycho 5", because they have certainly been controversial users on Steemit, and in my opinion (after having read through a sampling of their content, as well as articles from other users reflecting on the "pros" and "cons" of each of those psychos), have done way more harm than good for STEEM's potential. 2 of the Psycho 5 are the pair that we associate with "that" cryptoanalyst, and are not in the UA top 100. The other 3 are all witnesses currently ranked in the 30s-50s (though one is no longer an active witness) and show up in the UA top 100 (all with a UA score over 7). That by itself tells me that I should not trust the UA system.

But the other factor is that witness stake seems to be the "root" of all trust, which I find problematic. In an ecosystem as large as the STEEM blockchain, it is possible to be highly respected and reputed as a content creator but be completely unknown to any witnesses (especially in the case of building one's entire reputation on a non-blogging site/dApp, e.g. Musing.io). That means such a person would be subject to the luck of how well they fare in the proverbial game of "6 Degrees of Witnesses" to determine how high their UA score could get. And to allow 3 of the Psycho 5 to have such significant influence on top of that... that sounds to me like trying to fix a broken system (STEEM reputation) by using a different broken system (UA score). I think the latter broken system has to be fixed first before we can even think about bringing in freezepeach to enhance the fix.


I think your hardfork suggestions are a better solution that steem-ua, but I doubt it would be enough. But short of a system in which the blockchain can forcibly strip bad actors of their stakes (which I think has 0% chance of actually happening), I'm not sure what other measures could have a significant impact.

I partly disagree with you on this approach.

Dealing with abuses is one part of the balance. And I might be wrong, but I believe flag wars are only a fraction of abuses. I agree when there is no retaliation there's nothing from stopping the abuser from reaching higher highs. But retaliation brings in a matching retaliation (in time). I don't think many non-combative individuals will want to join a platform consumed by ever growing flag wars.

Do a full history reputation rollback on the entire blockchain

I doubt anyone would dare to try something like this. Many well-established steemians use or used at some point bidbots. I haven't, but I think this would provoke a major mutiny.

UA also comes with its own false curation

I also think @steem-ua gathered an important traction. It can and should be improved (and I'm thinking here about UA too). I think you are right about the false curation. But when true following and engagement is missing, what else is there?

Have @r0nd0n and his people moderate the flag queue. If they mark the flag, the UA service tries to cancel or attenuate the flag.

Give @freezepeach the voting strength of the UA service to use when needed.

I disagree that @steem-ua can be put in the service of @freezepeach. They have different roles, and most likely, support.

Replace the algorithmic curation with an algorithmic vote follow. That is, don't upvote posts of delegators based on the UA of the delegator, upvote posts upvoted by and downvote posts downvoted by the delegators based on the delegator's UA score instead.

I don't think it's possible, although it is an interesting idea. No single project would have enough SP to give significant votes in these conditions. Some bigger accounts (and even smaller ones) might cast a lot of votes per day at lower weights. The other issue is incentive to delegatators disappears or diminishes, and some are looking for it.

You got a 3.36% upvote from @postpromoter courtesy of @r0nd0n!

Want to promote your posts too? Check out the Steem Bot Tracker website for more info. If you would like to support the development of @postpromoter and the bot tracker please vote for @yabapmatt for witness!