How To Fix Steemit For Communities & Viral Engagement

in #steem8 years ago (edited)

Steem(it) needs to motivate bloggers to develop and bring in diverse communities of like-minded followers, which have the voting power to financially reward the blogger even without a whale’s vote.

Such engagement can lead to viral growth, because there would be a broader base of sub-groupings of interests served. Yet we must not enable the game theory vulnerabilities that were discussed in the following three blogs recently:

Solving this problem has perplexed everyone, including (originally myself and) the great minds of the creators of Steem.

Proposal For Fixing Steemit

I have contemplated a much simpler algorithm which seems to me to be more robust than those already proposed. That is we could weight voting power not only by STEEM POWER, but also multiplied by the ratio of uniqueness of followers relative to the system totals. In other words, if you have 10 followers that are not following anyone else, then you would have the same followers weight as someone with 100 followers who are also the followers of 10 other users. The community would need to contemplate for voting power what the relative weight between followers weight and STEEM POWER should be, i.e. the relative importance to assign to uniqueness (diversity) of community building.

Of course this algorithm could only apply to users and followers who are verified, otherwise it could be Sybil attacked. Verification could be a community effort, wherein a threshold reward in the #verification tag designates the verified. The community would police this #verification section.

Expected Benefit

What I expect this improvement to do algorithmically is spread out voting power more to those who have more readers’ attention than STEEM POWER. It motivates bloggers to not just go after the groupthink created around the gaming of the whales’ votes, but also gives them another motivation to create a following to drive more financial reward (since for example the system by default already has the blogger automatically upvote their own blog posts and presumably some of their followers might increase their followers weight as well and upvote the blog posts).

Without Vulnerabilities

The danger with other algorithms proposed for spreading out the rewards is that they hypothetically enable whales to game it by splitting up their STEEM POWER in multiple accounts, or by enabling a Sybil attack on account signups to be consolidated into powerful blocs, i.e. they don’t clearly reward productive behavior. But as far as I can see, my proposal doesn’t suffer these vulnerabilities because neither of those two attacks have any special advantage in attaining unique verified followers. My epiphany is injecting a valuable resource (the verified unique follower) which is concomitant with an activity (following and reading) that implies viral promotion and adoption. Note my proposal does not conflate curating with following, because it is not assigning a user’s followers to be that user’s curators. My proposal appears to be orthogonal to any pending decision to tweak curation rewards, since I am proposing a weighting modulation for voting power (although perhaps it could be applied to curation power if so decided).

Edit: I don’t expect my proposal to be without fault. This is to jumpstart the brainstorming process on some of the attributes I’ve identified.

Sort:  

The community would police this #verification section.

They would also have to police the selling of verified accounts (and I'm sure a market would be created sooner rather than later)... and that would be a nightmare / something very difficult if not impossible.

Good point, but even if selling of verified accounts is not policed, it wouldn't any worse than the power the whales have now. And those sub-communities that want to be functional and rewarded, need real followers, so it would still benefit them regardless. So I'd advocate not inferring with the free market, and do not try to stop the selling of the verified accounts.

These still need to be real people to achieve verification, so the value imparted to the voting power (over some reasonable time horizon) by a unique follower needs to be less than threshold of what they earn to become verified and this threshold has to be small enough that it isn't encouraging professional systems of hiring people to become verified who do nothing else but that, yet that threshold also needs to be significant enough that it isn't too easy to become verified such that is trivial to game it by having a bloc that upvotes #verification blogs (and thus not really be verified). In other words, the threshold can't be allowed to be so small that many can sneak under the radar of the community.

Another problem is that as the usership grows, people could reverify under numerous aliases and the community would probably forget these faces. But even so there is probably a limit to how frequently someone could fool the community. Typically the community rewards a comment post that exposes a fraud.

The game theory quagmire of distributing a pooled debasement of the money supply via voting won't be solvable without some community manual effort. Algorithmically it can't be solved purely automatically, because it is a Tragedy of the Commons self-referential given the externality of the shared pool everyone is trying to game. We have to introduce some external reference point.

The sybil problem is typically solved either by proof-of-identity, or collateral/investment. The proposal revolves around the first method with a "twist" about unique followers. I fail to see why this is necessary anyway. I mean if you solve the sybil problem, then you don't need second-tier algorithms anyway. You can just give a vote to anyone verified and let them run free with those votes. One person / one vote / system working as intended (and free of "inequalities" or sybil-attack).

The huge difficulty is on the verification side (easier said than done) plus not all people want to be verified. I would argue that there is a greater value proposition in even allowing anonymous posters to enrich the platform with their content - as sites like godlikeproductions do, although, apparently, we return back to the issue of "fair voting". I don't know - I'm not really a fan of verification to be honest for multiple reasons.

One person / one vote / system working as intended

Maybe, but influence as a function of SP is also intended, both as a significant part of the value of SP, giving a reason to buy it, and also because those who have made more of an investment (in contributions of valued effort and/or money) are more entrusted with influence. In that sense, the system might be working too well at the moment (or at least the whale critics would claim).

There certainly is a system where one person one vote is the intent, but this isn't it. Could that be changed? Possibly, but probably better to just build one with that intent from the start.

There certainly is a system where one person one vote is the intent

Actually I can't think of how to design such a system if the money is taken from a shared pool that is charged to the collective.

The only design I can think of which is one-user-one-vote and doesn't devolve into either a winner-take-all Sybil attack (which is even the case when votes are free such as on Reddit) or no curation at all due to optimum monetary strategy of voting for yourself always, is tipping from your own funds, which has been shown numerous times to not work psychologically.

You can just give a vote to anyone verified and let them run free with those votes. One person / one vote / system working as intended (and free of "inequalities" or sybil-attack).

Perhaps you didn't read carefully the linked blogs at the top of my blog or just missed it because there is so much to read, because it was already explained therein that one-user-one-vote (i.e. a linear weighting of vote totals for rewards) means everyone can vote for themselves as the maximum profit strategy and thus no curation happens at all.

The current voting algorithm is the square of the weighted vote totals. And this is to make sure people have to vote for posts that others also like. And rewarding (for curation rewards) the earliness of voting is so users don't get the same curation reward for just piling on the most voted posts late.

The huge difficulty is on the verification side (easier said than done) plus not all people want to be verified.

The community is already enforcing verification in #introduceyourself. And those who don't want to be verified, won't get higher voting power (same as it is now). Perhaps you are thinking that the bad actors can get verified (or buy accounts) and the unverified minnions will fall further down in voting power? Well if they are that unmotivated, maybe nothing is going to work.

I would argue that there is a greater value proposition in even allowing anonymous posters to enrich the platform with their content

That can also be a negative and chase away the masses, because they will run away if there is crime on Steemit. The anonymous posters can still post to the Steem database. Nothing stopping them from posting. We are just talking about rankings (that are optional for any UI to adopt).

I'm not really a fan of verification to be honest for multiple reasons

This is a social network, not Bitcoin.

The community is already enforcing verification in #introduceyourself

Somewhat. Mostly it filters out the weakest/cheapest frauds. And the vast majority of the users haven't ever done that (I think at this point the majority doesn't even post introductions at all, and many introductions are lost in the noise). Verification, in general, is an onboarding killer. Even for #intoduceyourself posts earning thousands of dollars there have been debates whether being "too strict" on verification is being too hostile to new users.

@alexgr:

Voting yourself can be rectified through relatively easy tweaks (making it unprofitable compared to voting others) so I wouldn't be too worried about that.

That is what the square weighting does. Linear weighting won't. You are making a non-point. Linear weight by definition means you can profit most by voting for yourself. If you tweak it, it is no longer one-user-one-vote.

Voting yourself can be rectified through relatively easy tweaks (making it unprofitable compared to voting others) so I wouldn't be too worried about that.

@smooth:

Verification, in general, is an onboarding killer.

Agreed.

I'm not diminishing your point, but I do want to note that if the users can choose to verify later at their time of choosing, it doesn't kill onboarding. Yet the new problem becomes why are they motivated to verify then if there is no urgency?

Btw, I had already thought of this. I will be making a new top-level comment soon to explain the disadvantages of my proposal discussed already and why I already know that we really can't fix Steem for as long as the model is voting to distribute from a pool of debasement.

Steemit became much less interesting to me with everyone worried about if someone is "real" or not over what they had to say. I think in some instances people will want to prove a real identity or ask for proof, but this making every newcomer post an ugly white sign makes it a lot less inviting and interesting to me than tumblr or wordpress.

Too much abuse of the system I guess that then made people ask for proof. But I agree - this is bullshit. Nobody should have to prove anything. Personally I just skipped the "introducemyself" altogether.

What if there were different reputations for different communities ?

Someone who has a good reputation from posts under the topic 'science' could have a different colorcode on his reputation than a 'makeup'-artist.

A steemit reputation of 50 solely based on philosophy posts could be valued higher in some circles than somebody else's reputation somewhere else ...

I hope that makes some sense

I'm a bit tired - I will post about it tomorrow.

In spite of the following, I think my blog is still important for the community to read, because we are brainstorming and discussing the very important design characteristics of Steem.

The important flaws in my proposal were pointed by @alexgr, @sunjata, and @smooth:

  • Verified accounts can be sold.
  • Desired connections between communities dilutes the metric for motivating unique communities.
  • Users don't have a strong enough motivation to be verified.
  • Intent of following is not a public metric.

I have deeply analyzed in my head (and various writings sprinkled over Steem and Bitcointalk.org) numerous facets and ideas for reducing the top-down effect of (the natural) power-law distributed STEEM POWER w.r.t. to voting for rewards taken from a pool of debasement of the collective, and I can't find any solution that isn't a degenerate outcome that rewards unproductive or winner-take-all behavior. The fundamental problem (design choice) is the rewarding of a collectivized pool of funds via a popularity contest. It is a Tragedy of the Commons. As far as I can see, there is absolutely no way to spread this out to motivate the investment in creating diverse communities.

We can observe recently the whales (or at least apparently @smooth) are spreading their voting out more, so that now the top trending posts typically have much smaller rewards. The only big paying rewards now seem to come when those whales (and perhaps a large contingent of dolphins) who vote less frequently, decide they really value a post such as Dogecoin creator's post today.

But that isn't necessarily motivating community building nor finding (ranking) content relevant to each user's communities. It is a top-down managed effect, and not driving diverse user actions w.r.t. to building followings. The follow and tags features aid somewhat these needs, but it isn't very monetary (mathematically the typical user can't earn more than ~$5 a month on Steem even if rewards of uniformly distributed) as compared to the reach a user can attain on a larger network that doesn't pay anything, e.g. Facebook.

The value of a social network is in the network connections and network size, not in the content. Users (even content producers such as bloggers) go where they can reach their audience (followers/fans) and friends (fans). There won't be viral penetration into the masses (excepting those ones who have a cryptonerd begging them to join Steemit) given there is no compelling reason for them to be on Steemit, i.e. the rewards paid out on average aren't compelling and the network is small with no incentives to drive network connections and no viral engagement paradigm.

In essence what I see on Steemit, are many people joining due to enthusiasm about the idea and thinking that others will too. They view their votes in the site-wide popularity contest as important for influencing this grand "we are changing the world" enthusiasm. It seems to me to be backslapping, but the masses won't see it that way. They will simply ask, "what is the advantage for me".

Edit: the above doesn't necessarily preclude other features and use cases for the Steem blockchain and token that might come to fruition. The onboarding gimmick of voting from a shared pool of debasement could be perhaps just enough to jumpstart an ecosystem, even though it might not (probably won't) virally spread to millions itself. I am skeptical though.

The following blog post quotes are confirmation to me of my idea that the relevance of community is essential to viral adoption:

I had bugger all Facebook friends, and most of them were “lurkers” (they never posted anything or even “liked” anything, let alone made any comments. So meaningful interactions were few and far between.

About half of what few friends I had left on FB disagreed with much of what I really wanted to post about, so if I got into subjects like conspiracies such as false flag shootings or “climate change”, or health subjects like low carb diets or the vaccination con, its safe to say I rapidly had less FB friends.

That giant pool of Facebook users don’t post much worth seeing, so to me FB was like a giant sponge that soaked up every creative urge in my mind and replaced it with a giant sea of shallow distracting crap.

The biggest reason I quit FB was that I wanted to see new ideas and be inspired by intelligent discussion. That wasn’t exactly coming thick and fast on FB

The following appears to incorporate some amount of "we will change the world" backslapping because he doesn't even mention having tried Medium for comparison:

Steemit has already amazed me with the quality of the content. I’ve only been aware of Steemit for 48 hours, but I’m already inspired by what I’ve seen here, and I’ve learned more in two days surfing Steemit than I did in the past two years feeding my FB addiction.

What I think is inspiring him is he sees some of his interests are shared by the community on Steemit. The Steemit crowd is probably into the specific idealistic issues he mentioned. And there are some smart Steemians. But I believe this sub-community can also be found on Medium (but perhaps not as highly charged and massive)?

Any way, the point taken is that communities and size of network (connections and engagement) within the communities is essential, which is my point also.

but it isn't very monetary (mathematically the typical user can't earn more than ~$5 a month on Steem even if rewards of uniformly distributed)

Maybe a small point but perhaps not since it relates to near-term growth. The current content reward pool is about $64000 per day. The current daily active user base is about 6400. If all of those posted daily and split rewards equally they would earn $10/day. That might be enough (even if not equally distributed in practice) to continue motivating signups and usage for the foreseeable future.

It is always helpful to have clear data. Thanks. I agree for that small size userbase. Perhaps you read my post at BCT where I made the point if they want to establish a merchant ecosystem, I think need at least 1 or 2 (and probably 3 or 4) orders-of-magnitude more active users.

One approach to gaming this is that in an open data system, "following" need not be public knowledge. I currently "follow" exactly no one using Steemit folllows. But I'm also following a group of people and I'm notified of their actions. If the former has significance to rewards then I can more or less shift around rewards at will (including accepting payments for it), independent of my actual interests.

Some method of identifying connections between people that are based on necessarily-public data such as having directly interacted via commenting or voting is probably better.

Agreed. Another example is I'm following some people not because they share the same interests as me, but because my interest in Steem at this time is more to study the dynamics of the system so I am observing some obscure example users to study their patterns of using the system.

It is very difficult to automagically measure groups, because as @complexring pointed out about the Grassmannian, everything is relative to everything.

P.S. see my reply to your other comment. Unfortunately Steem has a feature bug in that it doesn't notify of replies when the comment nesting is too many levels deep.

This is an interesting idea because recognising the value of people within tightly-bound parts of the network is really important. If you look at the theory of social networks, though, they are composed of both tightly-bound core regions and the "bridges" that hold them together.

In a Steemit-like social network, the people holding together the core might be user A who posts great photos of bridges and user B who posts great memes about cats. The "bridge" in that social network might be User C, who sees a photo of a cat on a bridge, points User C to a new tag, controlling the flow of information and growing the network internally.

Both roles - uniqueness and diversity - are really important for the health of a social network.

Both roles - uniqueness and diversity - are really important for the health of a social network.

Agreed. But it is difficult to think how algorithmically to separate those links between communities from one-size-fits-all popularity (that we have now) and following everything.

I am thinking that at least my proposal gives some power to those which have some uniqueness over the "everyone follows everything" crowd. So I think my proposal doesn't stop the diversity.

I think this is prone to sybil attacks since it probably won't be hard to create fake followings and game the verification. Ultimately Steem Power should be the foundation for voting and delegation of power should spread outwards from there.

I do like that you use followers as a parameter for delegation. I proposed a simple approach to delegate voting power to follower lists. This is natural because anyone you follow indicates: 1) the person is verified 2) produces content you're interested in 3) produces content you think others are interested in so you can make money on curation. All three reasons validate why those that are followed should have more influence.

The allocation could just be spread evenly among those you follow, so if a whale follows 50, voting power is split among 50 people. If a whale follows 1000, voting power will be split among 1000.

The other reason why this is good is because from a UI/UX perspective you don't want users to spend time creating their own delegation lists. It's easier to click the 'follow' button over time without much thought.

What do you guys think?

I would be somewhat careful about possible motives of following someone. It could be just to increase "social relations" or a way to indirectly flatter the followed so as to "reciprocate" the following. You can follow 5000 people in the hope that even if 1/10 of them check their followings and reciprocate, you get 500 followers back.

Well it's based on stake so if the Steem Power holders want to follow those using the followback strategy it's their choice. I think over time most people will avoid people who are using the followback strategy because there is actual value in who they delegate to unlike Twitter. The fewer people you follow the more control you have over where you're directing power.

I was just trying to make it easy and combine the follower/curation list, but if people feel strongly they want to keep it separate it would just be a matter of creating a separate list of delegates. I just think it's easier from a UI/UX perspective to combine it and for the most part the follower & curation list would probably be largely the same.

You can follow 5000 people in the hope that even if 1/10 of them check their followings and reciprocate, you get 500 followers back.

The uniqueness aspect I proposed defeats this.

But following is not a "state". It's in flux. I click the button now and I'm following you and then click it and unfollow you. And then click it again and I'm following you.

So in the scenario above, I get 500 reciprocal followers and then unfollow the other 4500 which didn't follow me back... lol?

So in the scenario above, I get 500 reciprocal followers and then unfollow the other 4500 which didn't follow me back... lol?

No problem. All this just get muted by the fact that everyone that follows everyone are no longer unique follows and thus deweighted relatively.

I think this is prone to sybil attacks since it probably won't be hard to create fake followings and game the verification.

Could you be more specific? How can they game the verification? The community is pretty diligent about fairness.

I do like that you use followers as a parameter for delegation.

Delegation is still a top-down control decided by whales. It isn't spontaneous formulation of new communities in a decentralized manner which I posit is necessary for viral growth.

It seems to me that offering some extra voting power towards motivating uniqueness of followed groupings, provides a balance between disconnected groupings and one-size-fits-all.

Over 60 rep you auto get up voted.
Than read after when have more time to enjoy the post. Keep it never fails me great content.

I like this idea, it makes a ton of sense to me. Valuing users more strongly who are attracting unique followers is important, however those followers should be weighted as well. Just weighting in followers who don't follow anyone else would highly value the creation of random accounts to just follow your specific accounts, and would bring new life into bots. Thus the SP of the followers should also be taken into account, to more heavily weight real people.

I like this idea, it makes a ton of sense to me. Valuing users more strongly who are attracting unique followers is important

Ty. Agreed.

Just weighting in followers who don't follow anyone else would highly value the creation of random accounts to just follow your specific accounts, and would bring new life into bots

No I don't think so, because they have to be verified and reach a certain threshold in their #verification tagged blog post. Which means they can't just be some random Sybil accounts. They have be convincing real users.

I think we want to weight by uniqueness of followers, because this separates out the sub-communities from the one-size-fits-all popularity groupthink in terms of where voting power and thus rewards go.

The whales benefit from the increased viral adoption thus driving the value of their STEEM POWER higher, meaning any cost to the whale in terms of reduced relative voting power weighting is I posit offset by unique diverse communities driving viral adoption.

Ah, I think I misinterpreted the point of the #verification tagged post. So the community decides whether an account is human or not. Nice. Okay, then I can see the complete value in this proposal. Hopefully the devs catch wind of it and can at least give it some consideration.

Hopefully the devs catch wind of it and can at least give it some consideration.

Yeah this is just part of the brainstorming process. I don't expect my idea is 100% iconclad or complete. Just throwing it into the mix of ideas.

Interesting proposal, however I'm not sure how having a 'unique verified follower' in and of itself brings value that warrants more influence. It might just mean the follower signed-up, verified themselves, followed the person that signed them up, took their introduction payout and quit the platform after a few posts.

On the other hand, there are people that sign-up, engage in the site, follow the person that got them to sign up, engage with and follow other like minded people and add value to the platform.

The more some integrates and forms bonds and friendships with the community/ a sub-community the more credence it gives to their verification.

Also, under this proposal I can people paying individuals they know (who have no interest in the platform) to sign-up, verify themselves and follow their account.

It might just mean the follower signed-up, verified themselves, followed the person that signed them up, took their introduction payout and quit the platform after a few posts.

Then they aren't using their higher weighted voting power, so no problem at all. Their voting power is returned to the community which is active.

On the other hand, there are people that sign-up, engage in the site, follow the person that got them to sign up, engage with and follow other like minded people and add value to the platform.

And presumably these are the ones that continue voting as well.

I'm late to the discussion, but if you have verified users, why wouldn't you use pagerank instead of your ad hoc metric? Is it because you want to avoid pagerank's n^3 complexity?

In addition to the n3 complexity, would pagerank incentivize bloggers to recruit followers from outside of Steem and introduce them to Steem? Isn't my proposal more explicit to the blogger as to what they need to do in order to increase their voting power? Nevertheless my proposal has flaws and I don't think it optimally aligns the incentives for viral promotion and adoption.

Coin Marketplace

STEEM 0.28
TRX 0.12
JST 0.034
BTC 63152.69
ETH 3233.54
USDT 1.00
SBD 4.45