RE: Why I Advise Against Linear Reward
the biggest shareholders have a disproportional incentive to defect but if too many do, they collapse their own network and Steem net worth. Thus there is an incentive for them to police each other
This part is totally bogus. There is no body of 'biggest shareholders' and most if not all generally believe they can gain personal advantage and not worry about the whole. That is human nature and demonstrated time and time again including on Steem.
Possibly if you had a much higher actual stake concentration of one or a few truly dominant stakeholders AND it were stable AND there weren't ways to cheat the cartel and get away with it (all false) then it might be valid. But in reality it isn't even that. We have whales with maybe a few percent of stake at most which are far to small. Even whales for practical purposes just end up being regular users most of the time, albeit with bigger votes.
There is no meaningful difference in incentives to downvote when someone has 1% stake holding vs. 0.00001% stake. In either case it is done almost entirely to express an opinion with limited to negligible direct economic incentive or impact. Both may also feel some sort desire to do good for the platform, but it is primarily emotional and not economic, in both cases.
Continuing to quote the white paper on this is entirely boring because experience shows otherwise. When theory conflicts with data revise the theory.
The only benefit I've ever been convinced of from superlinear is reducing dust farming, which to be fair is not nothing, particularly given the lack of any robust incentive for any individual to care about or take action against abuse (see above). The other arguments don't hold up.
Also, the 'defect' terminology really doesn't make sense here. There is no clearly defined cooperation criteria. A system like this is too large and complex and too subject to differing opinions and attitudes over what constitutes mutually beneficial cooperation (as well as the nature and magnitude of any benefits to be achieved therefrom) for cooperate-vs-defect to be a useful label IMO. If people can't even agree on what it means to cooperate, they can't decide whether to cooperate or defect. The entire model is inapplicable.
I stand behind the original whitepaper's argument. We disagree. My commentary made on my original post has inexactitudes and I plan on possibly strike them through.
Dust farming is vote farming. I don't see why some votes should be left out.
This is also true for superlinear rewards but under superlinear rewards, consensus influences the distribution of rewards and flags and thus every vote and flag influence one another which isn't so under linear rewards.
In the long run, if the platform, Steem price doesn't perform well then the biggest holders lose more from gaming the system then the price drop.
Getting paid for 0 work is the only detrimental "work" being done. Any work however controversial has value. To defect is to create 0 work i.e. exploit a loophole if one exists and abuse it as long as possible at the detriment of the whole.
As the price increase, the incentive to defect (i.e. self-upvote) grows and flags don't solve this as the flags simply return the reward to the reward pool, making it bigger proportionally to the number of people remaining eligible for it.
As a whole, the smaller the voters, the easier for them to defect without being caught.
I'm still very much enjoying this conversation as I think it's very much important and it really seems like you care even though we might disagree in the end.
Then you are either a fool or irrational. To continue to stand by a theory when observed data don't agree is not the basis for any conversation I care to have.
At a minimum it needs: a) a reasoned explanation for the discrepancies, and b) additional qualifications and refinement to agree with new observations.
Come up with a theory that is predictive and also explains observed behavior. I will be happy to discuss it. The refuted white paper theories not so much.
Everyone does. There is no special role for 'the biggest holders' here (even if that term were defined, which it isn't).
Let me clarify that. Superlinear reduces the quantity of content that people need to consider downvoting. That's beneficial as a labor saving device.
Provide a formal disproval of what is in the whitepaper. We disagree on the data but you haven't provided any holistic data.
Someone losing millions of dollars from Steem price drop does lose more than someone who loses only a couple of dollars. They lose the same proportion but that's not what I was arguing.
I stand by those who say the whitepaper theory hasn't been properly refuted. I hope you can respect that. This is my honest conclusion from my limited human mind.
We had a year of experience with n^2. The theory that the largest stakeholders would police each other was shown to be false. People were shown to be more concerned with seeing to their own rewards than spending valuable vote power downvoting others.
To complement that observation with theory, the economics of being more concerned about the value of their steem investment that personal gains don't hold up. For any stakeholder (at least <50%) the personal reward is greater than the share that hurts the value of that stakeholders share of steem's total value. The rest of the cost is borne by others. A system with these tragedy-of-the-commons properties is badly designed and needs refinement or replacement.
The "large" numbers are irrelevant and are being thrown out for scare value. A large stakeholder may care less about millions of dollars than a small stakeholder may care about couple of dollars. You can't generalize. Furthermore the incentives to self-reward scale accordingly or in the case of superlinear, more than accordingly. The one at risk of losing millions can gain thousands or more per vote. The one at risk of losing two dollars can gain either proportionately or less. There is no sounds argument here that supports large stakeholders being 'good shepherds". None. If you think there is, please make it, step to step, and showing specific incentives (in at last a qualitative sense) without resorting to scare numbers.
Most people care about their investment. Even Warren Buffet cares about his investment. Maybe a minority don't but on a whole, they care. If they don't care we can't predict what they will do anyway, regardless of the economics.
You're talking here in terms of the number of people. Someone who has 1 million Steem is more vested than someone with 1 Steem. They don't have the same amount of $ to lose, not the same vested interest.
On Steem the incentives are counted in Steem. If people don't care about their Steem it's futile to try to predict their actions based on economics. So the majority shareholders have more to lose and should according to behavioral economics care about the net worth of the Steem network. Their own net worth health first pass by the value of the Steem rather than their reward.
Yes of course but this applies at any size. Someone who has worked his or her ass off to accumulate two dollars worth of SP may care about it more than a while with a million dollars worth. Don't be blinded (or attempt to blind) that "large" dollars figures somehow automatically convey some more sincere or "better" intent. They don't.
Sure, and everyone can care or not care about their Steem. The amount being 'large' doesn't inherently change anything, except maybe how YOU perceive it.
That doesn't follow. It can apply (or not) equally to stakeholders of any size.
I've used the term more vested interest.
The amount matters. Different amounts will be perceived differently for people but what people value they don't want to lose.
Could you re-phrase that?
Someone with 2 STEEM may very well be as concerned, if not more concerned, about its value dropping than someone with 1 million STEEM. Pushing all of the influence over rewards to the latter an then relying on them having 'more to lose' from the value of STEEM decreasing is nonsense that does not work, both in theory and in practice.
There is no incentive of good stewardship that scales with the size of stake, as the whitepaper suggests (as do you). It's a false model.