Diversifying Curation RewardsteemCreated with Sketch.

in #steem8 years ago

In my previous post, I suggested to convert curation reward into investor reward, which is given when Steem Power is locked and unused. There have been backlashes on this and as far as I understand the main concern is "discouraging vote while encouraging doing nothing". This post addresses the concerns and reports revised proposal with detailed numbers.

Rewards for Nothing?

It seems fair statement that giving rewards for laziness makes no sense. However, are they really doing nothing? No. To have Steem Power, they bought or earned STEEM, powered up, and have been holding Steem Power. If you say holding is doing nothing, you may prefer selling and are enjoying recent STEEM drop. Although holding doesn't actively add value on Steem, it is a large supporting factor for STEEM price and at least doesn't decrease Steem value.

Congestion Reward

In normal case, some governments charge congestion "fee" when the road is congested, so that they discourage traffic with lower willingness-to-pay. But what will happen if governments subsidize $5/mile for driving freeway? A rational choice is driving as much as possible to get the highest rewards. I think this is very similar to what we are now doing. Currently, curation in Steemit requires low costs. You don't need to pass CAPTCHA, don't need to read, or even don't need to open the browser and access to Steemit. If you want to minimize the costs, using bot with good strategy is the best option. Even though you don't have good strategy, you are still earning more than just holding Steem Power. You drive anyway not to relatively lose money. Many Steemit users upvote posts to earn while others do too.

Potential solutions

Since profit consists of benefit and cost, there exist two approaches to resolve this problem. As cost side, increasing costs for vote is a clear way. If people needs to spend their resources(e.g. time, money, effort, etc.), part of voters with lower willingness-to-pay will give up. Unfortunately, this is almost impossible with blockchain-based system like Steemit. Voters can directly access to the database(Steem blockchain) and upvote without any hassle.
Regarding benefit side, the most radical solution is merely removing curation reward. As I mentioned in my last post, "psychic income" only remains as a benefit so voters mainly with financial motivations will quit. However, this solution would decrease overall demand of Steem Power as a side effect. Therefore, as the third way, my suggestion was creating investor reward to encourage holding Steem Power. But I have to admit that I was not careful enough with suggesting numbers. Zero curation reward may cause negative perceptions, and it is also true that it discourages voters with low psychic income but still have it anyway.

Having Both

To suggest specific numbers, my question was the degree to which stakes much care about financial incentives and how much does not. Investigating bots doesn't help because humans also can vote for financial incentives. A better way maybe listing accounts that have relatively poor performances in curation rewards. For instance, as of 2017/2/21, proskynneo showed about 1/5 curation rewards compared to wang while they have similar amounts of Steem Power. Actually, proskynneo's reward is smaller than dunia who has 1/10 level of Steem Power. So I assume that proskynneo does not much care about curation rewards. Using this method(and great help from @liberosist), I found that about 20% of total Steem Power (and about 40% of active voters) falls in "I-don't-care" category (about 90k MVESTS). They consist about a half of rshares and curation rewards.

More specifically, for-profit voters, who are holding about 30% total stakes, take approximately 10% of total content reward(author+curator) and for-content voters with 20% stakes takes other 10%. This means, to satisfy for-profit voters with investor reward, we need to provide at least 27% of total content reward to them. For-content voters can have same level of reward if we assign 10% of total content reward, so that they have no changes.

To give relatively more incentives, we may want to increase investors reward more. So I think 35~40% needs to be given as investor reward and 5~10% for curators. An issue is smaller proportion of author reward. A possible way to resolve is allowing to decline investors reward and excluding steemit accounts that have about 40% of Steem Power from investors reward. This makes investors reward doubled, so then we only need 18~20%. Adding curation reward of 5~10%, this will take 23~30%, reducing author rewards only 3~10 percentage points.

Open to Discussion

This suggestion is for opening discussion. The number are roughly calculated so probably not accurate or even totally wrong. Any feedback, correction, and criticism are more than welcome.

Sort:  

tl;dr

just joking... actually I read it twice and still have NO (real) idea what you are talking about. Especially it seems to me that your proposal does not solve the problem of self voting by the guilds, which is mainly distorting the for-content voting they are involved in (or should be).

Still glad to know, that there are people doing the maths ( I am not fit for that ).

At the end it will be key to finally find a stable solution that is not subject to change all the time. Theses changes are one of the main reasons investors are hesitant...

Hoping for a lot of discussion here, to learn more about the maths behind the system.

At the end it will be key to finally find a stable solution that is not subject to change all the time. Theses changes are one of the main reasons investors are hesitant...

Exactly. And they are also hesitant to "invest" when their only reasonable "returns" are continually being stripped away without producing the claimed results - which are usually the promises of more users, more investors, more development, and a better overall experience on the platform.

It is not directly related self-voting. But notable thing is these for-profit users/bots don't much care about these self-votings. If a post will give them higher profit, they will vote for, and exacerbate the problem.

You seem to be reasoning from the assumption that automated curation is necessarily a bad thing. Think again and look at how things already work in the traditional economy.

Subscriptions for example. What are subscriptions if not systematic voting and rewarding of content over an extended period of time? If I have a subscription to Wired, does that mean I am endorsing all and every article of Wired? Or are some articles complete crap and still being rewarded by the fact I maintain my subscription regardless because I feel overall their content is good? When I take a Netflix or Spotify subscription, is only the content I actually like being rewarded, or is all the content I consume being rewarded regardless of whether I liked it or not? Does that mean that subscription-based models are ruining the content production business in the traditional economy by destroying incentive to produce quality? No, because subscriptions typically only go to producers who are consistently creating good quality content. Of course there is some latency due to the commitments and the time it takes for reputation to wane, but eventually content producers whose content start to disappoints will lose their subscribers over time.

I see bot voting exactly as subscription-based content consumption. I have a daily budget that I have to spend (the voting power) that comes as a perk for being an investor in Steem. When I have the time, I can decide case by case how I want to use that budget to promote content I like or think is worth keeping on Steemit, and if I don't have the time, I can identify authors which content I find consistently good or valuable for the platform and support them over a period of time in the hope that they will continue producing good and/or valuable content. How is that a problem?

What I have observed so far is that people who used to be top performer and started getting lazy after falsely perceiving that they had unconditional bot support ended getting dropped from bot lists after a while. Of course, it takes some time for things to adjust. But eventually, all it takes is for the actual human reviewers to stop voting or downvote or make noise and complain to create a beginning of trend reversal of the author performance. Soon enough, the content starts to become less performing and even systematic bots that involve no human intervention will eventually detect that and move on to more rewarding content.

Bots that are only reflecting their owner's general opinion turning case-by-case consumption into a subscription-based consumption are just fine in my opinion and only mirror the way things are already working in the traditional economy.

As for bots that just follow trends set by other people, they are just amplifying the trends set by actual human reviewers and making the rewards exponentially rewarding for the best authors. This only reinforces the effect Steemit already deliberately engineered by making the reward grow superlinearly to the amount of stake-weighted votes received, which gives a little "American Dream" side to Steem that isn't without its appeal.

I think the problem is one of balance. Although I think all types of voting have their place in the ecosystem, it is paramount to make sure that bot-based amplification doesn't amplify noise. For that, the curation signal has to be strong enough, and I think that the "problem" you currently perceive and try to address is caused by the imbalance between bot-based voting and manual voting, which itself is caused by the extreme imbalance of Steem Power distribution and actual demographics involved: all the Steem Power is between the hands of crypto-nerds who are also the most likely demographic to be too busy trying to keep up with the fast paced evolution of the field to spend significant time reviewing content on one of the many projects they are following and/or contributing to. This is my case and I think the case of most other botters.

Is the fact this demographic is getting the lion's share of the curation rewards a bad thing? Well, yes and no. It's a bad thing because of the current noise amplification effect. But in the long term, it's a good thing because this is precisely the type of people you need to grow the ecosystem, develop new businesses, and convince other stubborn crypto nerds that Steem is the real deal. Their contribution may not be so visible in terms of content and curation, but their effort in the backstage where a giant part of the battle is being played is significant. Without this demographic, Steem would fade into irrelevance as a crypto project.

Luckily, as the early adopters of Steem are powering down to finance their projects and other efforts in the crypto sphere, everyone is getting the opportunity to buy large amounts of STEEM on the cheap. Steem is near its all time low. Now is the best time to promote Steem to a more diverse demographic and make sure enough of them gets a stake before Bitcoin whales start pumping the price.

It's not emphasized here but on steemit.chat and in couple comments (here and here) I clearly stated that bots are just automated humans, so they are neutral. I agree with you.

The point I highlight is a motivation of for-profit users(including both human and bot), and incentives that form the motivation. Steemit system gives incentives to voting, but the cost of voting can be lowered near zero, especially when using bots. While curation by human costs significantly more. These are facts I start with.

If one's purpose is maximized financial incentives that Steemit gives to voters, his/her rational choice is lowering cost and using voting power as much as possible. In Steemit, lowering cost means no reading, no evaluation, no decision-making. Maximizing voting power usage means use about 40 votes(at 100%) a day. There are also curation auction things. Bots seem problematic because they are best-fit option for given the system.

Unfortunately, as you know it is almost impossible to increase the cost side since anyone can access to the blockchain DB and make transactions. Then other is reducing incentive, e.g. removing curation reward. But this is quite radical and consequently decrease demands for Steem Power. An alternative way I suggest here is giving some compensation to for-profit users, not in curation reward that requires votes, but in other types of reward. Investors incentive is an example to detach for-profit ones from for-content curators.

Thanks for your long and detailed comment. During discussion, I realize participated people already added values to Steem. I really appreciate your and their time and energy for arguing, regardless of supporting or opposing.

If you agree that bots have their place, and are basically automated humans, why the emphasis on trying to weed them out by fiddling with incentives?

You are claiming that voting, as a bot, is significantly less costly than it is to a human. Right now this may be true because the market is still deeply inefficient due to it being very young and some very simple voting strategies are consistently working and affecting negatively the market (which is why bots are not seen in good light right now). But keep in mind that in a competitive environment (not the case right now) bots too need to make decisions to maximize their profits, and these decisions can't be made out of thin air. Either the owner of the bot regularly seats down and manually curate his bot list to weed out underperforming authors, add new rising authors, and identify potential new talent. Or the bot is fully automated in which case it needs to take its input from somewhere, this somewhere being all sort of things from past performance to keywords based heuristics to simply following someone else whose performance as a curator has proven to be good. Now, we aren't yet in an era where programs can parse and understand natural language, so once quantitative tricks start yielding too low margins due to increased competition, botters will have to move to qualitative analysis of the content to determine what content is actually valuable as this is the best predictor for success and higher returns. And what better way to do that than hiring curators (like Smooth is already doing) or following successful curators? Because of that, being a human curators with a talent to pick good content will increasingly lead to accruing a significant bot following and in some case land job opportunities as exclusive curator for a large stake holder. This will increasingly incentivize curators and redistribute some of the wealth captured by bots back to human curators. This in turn should serve as incentive for human curators to spend time reading content and building a track record of good picks for the bots to detect and act upon.

Why do I know this will work? Because this works exactly like that on the traditional financial markets. There are basically two main categories of trading: fundamental analysis and technical analysis. Fundamental analysis involves analyzing the assets in their economic, societal and political environment and determining based on their intrinsic qualities and fitness, as well as the image they project in the market, which assets will appreciate and which will depreciate. Technical analysis knows nothing of the assets and only analyzes market behavior, using all sorts of statistical methods to compute correlations, trends, momentum etc. to attempt to determine what to buy and what to sell based on mere probabilities. And at a higher level, there are traders who buy research from technical and fundamental analysts, consolidate the signals, and trade based on a hybrid model that combines both views. These days, there is even a new approach called "social trading" that involves specialized social networks for traders where traders both institutional and individual can propose macro views, analysis, strategies, signals or pronostics and get rated based on how much predictive power their research turns out to have retrospectively. The best rated "social traders" are being increasingly followed with the top performers being routinely poached by leading financial institutions. And then you have social traders who trade using optimized baskets of signals from other "social traders" etc. The point being that social trading is the ultimate melting pot of trading and AFAIK the best present manifestation of man-made swarm intelligence, and yet ... social trading involves both technical and fundamental analysis and anything in between, to the point where the limits is becoming completely blur and irrelevant.

Of course, at some times, it seems that systematic trading is taking over the market, and you will hear the fundamental analysis folks and other economists (who are generally having the biggest corporate titles and shout the louder) cry that bots have seized the market and more regulation is needed. You always hear something like that after a flash crash where it turns out that a bunch of dysfunctional bots raced the market to the ground. But in reality, fundamental traders are still completely driving the market. Bots hardly take any direction at all, at least not on a macro level. Bots have no emotion: no fear, no greed. They just find alpha and take it. The important market movements, bubbles of all sorts, rallies and corrections each time the Fed scratch their nose, market hickups on a political events and corporate jetset gossip, hype cycles, bear and bull markets etc. all this is the work of fundamental traders. They are the ones who move the trillions between markets and asset classes. Bots just smooth the curves and arbitrage the little inefficiencies here or there, detect and amplify the trends, middleman market movements they anticipated and accelerate mean reversion. Bots are just swarming around the real decision makers of the market, picking the eventual bread crumbs and improvising themselves insiders of actions they don't understand. In spite of all that has been and is still being said on high frequency trading and automated trading in general being evil and the cause of all our troubles, bots have never, do not, and will not before a long time be the ones responsible for moving the market in any direction whatsoever except for the occasional bot accident.

Now, how does that relate to Steem? It's simple. Curation voting is trading in disguise, with manual curators acting essentially like discretionary fundamental traders voting based on qualitative analysis of the value of the content, and botters being systematic traders trying to derive signals from the technical analysis of the curation reward history, and modelling of the behavior of manual curators and other bots and the patterns their actions. Right now, I concur that there is still a strong imbalance between botters and curators, with botters having almost all the trading capital and overfitting a very very weak signal which lead to them amplifying mere noise. But with the recent price movements that lowers the bar of entry in the "whale club", the new budget allocation rules that favor content creation in the redistribution of inflation, and the fact early adopters have been and are still continuously powering down, the balance is slowly but steadily shifting toward equilibrium between value minded curators and botters.

As the curation signal gets strongers, bots will actually start adding value, for instance by making the market much more responsive to the discovery of new talent. Right now, a promising new author may hover at mid-level in the Hot section without making it to the Trending section in spite of numerous small votes from real users. With more and smarter bots, the good lexical footprint of the post, together with the fast increase of popularity of the unknown post from many small accounts among which some may have good and uncorrelated curating track record will immediately signal good content and attract bot votes, propelling the new author well into the top tiers of the Trending page. This is just one example of how bots could eventually turn out to be the best thing that ever happened to Steem in retrospect.

The bottom line is that trying to prevent bots is both counterproductive and pointless. Bots are an unavoidable and actually beneficial element of a well structured free market, the emphasis being on "well structured". Instead of focusing on the sisyphean task of getting rid of bots, energy would be better spent on pushing for a higher diversification of stake ownership either by promoting more heavily Steem to demographics that are likely to provide good content creators and curators (college students, academics, artists, authors, columnists, bloggers, web marketers, ebook writers etc), campaigning to convince existing content creators and manual curators that holding their SP / powering up their liquid earnings is a good idea (which would provide a refreshing change after having read so many posts from authors who complain that stake is imbalanced but are powering down their holdings anyway hereby making the situation even worse), and last but not least lobbying SteemIt Inc to start doing carefully designed and thoroughly KYC'd Steem Power giveaways from their gigantic premine to the same hand picked demographics I mentioned earlier.

bots too need to make decisions to maximize their profits

I think this can be a start point to explain why you may not understand my argument. I fully agree bots need to maximize their profits as like humans do for utilities. But the question is here, what is required to maximize profits for bots (or some humans)? As you mentioned we haven't reached enough to where bots can understand and feel contents in natural language, they mostly vote with quantitative measure. Not only bots, some humans can follow some authors and every 6 hours mechanically upvote all posts on their feed without reading them.

You may think differently, these are mainly generate much noise in terms of qualitative curation based on content. For instance, if I post consisted of random words (e.g. "feawf 23`1908j f;adsofsda") and self-upvote, there will still be tens of upvotes for that post, and probably several dollars from my experience. For many users, especially new users who don't fully understand the patterns of our system, these votes are just noise that hinder efficient content discovery.

But the question is here, what is required to maximize profits for bots (or some humans)? As you mentioned we haven't reached enough to where bots can understand and feel contents in natural language, they mostly vote with quantitative measure. Not only bots, some humans can follow some authors and every 6 hours mechanically upvote all posts on their feed without reading them.

The whole point of my post is that it is a normal and actually a desirable thing in a healthy free market to have second order (meta level) agents who do not make decisions based on the perceived value of the asset itself (here content) but on the behavior of other market participants.

Please read more carefully my previous post. I think the analogy with traditional finance should make the unavoidable existence of these two levels very clear and explain why having bots that do not understand at all content is not at all a problem (and actually a good thing) so long as there are enough participants in the market who do, which can only be achieved by diversifying ownership of Steem Power, not by trying to fiddle with the rules to weed out bots while keeping human participants who vote based on perceived content value.

You may think differently, these are mainly generate much noise in terms of qualitative curation based on content. For instance, if I post consisted of random words (e.g. "feawf 23`1908j f;adsofsda") and self-upvote, there will still be tens of upvotes for that post, and probably several dollars from my experience.

This isn't bots generating noise. This is you generating noise and bots amplifying it because they don't have any other clear signal to amplify and work on the very basic (but historically consistent) assumption that you are a reliable curator and author. If you keep creating nonsensical noise, eventually bots will adjust (either automatically for the most advanced ones, and manually for the managed ones) and stop following your votes and/or upvoting your posts. Granted current generation of bots is pretty dumb, but again the reason is that there is very little competition and it's fairly easy to reap a good curation reward with minimal research effort. As Steem becomes more popular, there will be more bots and more manual curators, so there will be more competition and a better incentive for bots owners to up their game and make more effort discriminating actual signal from noise. Paradoxically the best way to get rid of over-simplistic bots is to encourage botting so that more people will join the game and push the bar higher.

At the risk of repeating myself, you won't turn and immature market into a mature market just by trying to rein in the bots. As you said (albeit not in these words), bots are automated human thought processes. So long as the system will offer economic incentives to humans based on the completion of tasks that do not require self-awareness or involve emotions, it will also by the same token offer the very same economic incentives to bots.

We could remove entirely the economic incentive and make the reward entirely emotional so that it would make the use of bots pointless, but then you are back to the traditional blogging model where people post to share with their peers and gain social mileage, and even that is plagued by bots! Just look at all the click farming for "likes" and "friends" and "followers" and cheer leading comments.

Without economic incentive, the game theory that underlies distributed consensus can't exist. Without sound distributed consensus, you open the door to sybil attacks and manipulation. Back to square one.

Loading...

Incidentally, as what might be possibly an example of bots amplify bot signal echoes -- check out this thread:
https://steemit.com/f17/@got/combust

Im testing something for a site im working on. If you look at my comments, they seemed to be getting 10 bot votes yesterday, but tonight are getting 15.

Thanks for your points and I agree many of them you made. But I think debates on bot is not a main issue (not even in the title) regardless of their smartness or dumbness. The point is design of incentives.

Numbers can be inaccurate, but for example Steem is giving 0% to whom don't vote and 1~5% ROI to those who vote. So voting is always more profitable than not voting. Rewarding on participation seems good, however, there is no alternative to keep profit other than vote.

The incentive system what I try to design is like this. Steem Power basically has fixed ROI of 2%. If you vote, you will get between 1~5% based on your voting performance instead of fixed 2%. If one is very confident to vote well, s/he will dive into the voting game, but if one prefer more stable income s/he will stay on 2%. That's my main idea of diversifying rewards.

Regarding this, I wish steemit accounts are excluded from fixed income as well as curation rewards because they are not an investor.

Coin Marketplace

STEEM 0.19
TRX 0.15
JST 0.029
BTC 63630.04
ETH 2656.44
USDT 1.00
SBD 2.81