DPOS and Dunbar Number: Can People Manage 30 Witness Votes?

in #dpos4 years ago

Image credit

A major criticism on DPOS is its small size of producer group, currently 21 in Steem and 25 in BitShares. As @iang mentioned, "How large is large" is always the question and balancing between more decentralization with inefficiency and more efficiency with centralization is hard to achieve. However, given that centralized hashing power distribution (e.g. Ethereum's top 20 miners consist 99.4% and Bitcoin's top 20 has 99.5% of hash), our opponents are not free from this "smallness" issue as well.

On the other side, there is another criticism that DPoS gives too many votes larger than a producer pool size (e.g. 30 votes for 21 witnesses) and therefore it opens collusive voting for whales and centralizes power. Technically, this is not true since whales can split stakes with small number of votes. However, I think this still matters from the perspective of cognitive burdens that humans can deal with, so-called the Dunbar's number.

Generally, Dunbar's number is known as 150, which human can maintain relationship comfortably. But in his article "Discrete hierarchical organization of social group sizes, various sizes by hierarchy was suggested. More specifically, people can have 30-45 of "close friends", and 9-15 of them are "intimate friends", and 3-5 of intimate friends can be called "best friends".

Then, what does 30 of Steem witness votes mean? It means you need to use third level of cognitive burden. To manage and closely monitor 30 witnesses, voters must pay greater efforts than have 15 or 5. Surely, people can only vote for few witnesses. However, it is also true that people can vote 30 times without prudence, or even they can use the leftover (meaning they already voted for whom they value but there are more votes) for vote trading.

There should be additional incentives to differentiate active voters who are willing to use their 2nd or 3rd level of cognitive energies. However, this incentive must not exist too much. In my opinion, a better decentralized governance can be achieved when as many citizens as possible are acknowledged and act in favor of the platform. For this, I suggest to reduce witness votes to 5-9, which is in between the thresholds of best friends (3-5) and intimate friends (9-15). This change will make active producers into multiple groups according to voter's interests. With the smaller number, voters are able to pay more attention to all producers they are voting and more actively response to producer's performances/stances.

Relative posts:
Seeking Consensus on Consensus - DPOS or Delegated Proof of Stake and the Two Generals' Problem
DPOS system for Steem ?


Very good. I am fairly new to Steem but the first concern I had over voting was that the number of votes is greater than the number of witnesses. This means that with voting collusion only 50.01% of token holders can collude to 100% dominate the witness landscape. If the votes granted were only a fraction of the number of witnesses (as you propose) then it is harder to dominate witness positions via whale collusion, leading to a richer, more diverse ecosystem. Factionalisation will occur over time - it is inevitable - but if the democratic foundations are stronger then the community will be better placed to have the required robust debate between diversified witnesses when the time comes. It is unrealistic to expect regular users to utilise 30 votes intelligently, but 5-9 is very realistic. I hope you get some traction.

Dunbar's number applies to relationships and "friends", I think that most people can take queues from others and so the load for selecting producers is not the same as the load of maintaining an acquaintance. Also, those who are more active will have more say than those who are less active by virtue of their ability to vote while others do not.

The key of Dunbar number is "maintaining" relationship. There are some votes unmaintained at this moment on Steem (e.g. inactive witnesses still have 10 mil SP votes). And if people perceive most witnesses are determined by few whales, voting indifference will arise.

It's not clear that there's any relationship between the cognitive cost of maintaining an interpersonal relationship and the cognitive cost of determining the acceptability of someone as a witness.

Voring for witnesses is also interpersonal relationship that delegates my power to trustable entities. But it is not one time event. People need to spend some psychological energy to ensure delegatees act properly according to voters' interests.
The quality of relationship must differ but this maintaining delegation may have less paychological resources than personal relationships.

Can average people monitor 30 witnesses very carefully? I don't think so. They maybe able to do less than 10, 33% of max votes. Although I do not disagree that active voters should get incentives, I think 67% loss is too big.
Similar discussions were on max upvotes per day, i.e. 40 vs. 10 (current). When too much resources are given that human cannot handle, side-effects are more likely take place.

I don't believe they need to monitor very carefully. Instead of close monitoring a witness, people can keep an eye on their friends posts and the trending page. More dedicated users will read witness-category. If a serious issue is raised about a witness, the event information will be propagated from person to person.

While imperfect, this is good enough to allow people to update their witness votes, as long the voters see it as important in the first place.

I agree your statrment, but only generally. Cognitive costs can differ between friendship and witness support as you said. Paying more attention to friend's posts is more important too. But do they directly justify 30 witness votes?

My point is the number is too largr to the majority of people can handle easily. Because their priority on Steemit should be to communicate with their friends. Be advised that abusers are also dedicated in a negative way. Giving too much merits to exteme sides seems not alwaus a good idea.

Approval voting best reflects group consensus when you can vote for as many people as you wish, that is the justification for a large number of votes. The only reason it is limited to 30 votes is to reduce the computational complexity (and thus potential DOS). It would be justified to have infinite votes otherwise.

very good.
I would be happy if you like to follow me and give your opinion about my posts.

Humans Nodes have limited scalability in relation to the number of interactions with other humans and looking at all species is proportional to the size of the neocortex.
Recent studies of Dunbar and Al. found that there were layers related to the strength of emotional ties with peers (5 more close friends or family in the first layer, then 10 less contacted friends in the second layer, then 35 and a final layer of 100, defining a stratified nature for human social interactions.

Dunbar's numbers simply say that the value of a personal network, is around 150 and maybe can reach 500 but after that is surely very difficult to manage, even with the facilitating mechanism of the Internet, because each connection demands a finite amount of time.

The importance of this numbers here is to understand the capacity to keep people feeling human on being equalitarian, to access and influence and I will call it the Human value of a social network that have always to be analyzed in therms of culture and social stratification of that community.

In a network node connections in the beginning increases but around Dunbar's Number (150) some data has showed this values can until until 500, varying a lot with factors of age and time spent online, but only a small percentage of those "friends” actually communicate with one another.
The number of stable of social connections a person can have, overstating network value is ignoring peoples capabilities of the individual value to form networks or communities bigger than Dunbar's Number or as other studies can go from 150 to a maximum of 500 depending on age and online time engagement that can may be helped and supported in the future with new filters, attention economy layers and new tools.

Group sizes larger than this can only became hierarchic by imposed rules, laws and regulations to maintain a stable but coercive cohesion.

I definately support your view on an additional incentive for the active members who are willing to use it.
Pls upvote my post

Very informative, thanks for sharing! Upvoted and now following.

Rule of 150.. even at my current workplace, I was employee #17. Now there are 60+ people and the I can feel that the social vibe has changed drastically. Voicing my opinion has less meaning now. I believe I had the best time and productivity when we had about 20~40 people. But that's work place. In this social network type of platform, people would care less about importance of the votes. I do concur that 5-9 witness votes would increase the quality of the relationship even further.

nice post thanks for sharing...upvoted and followed

Very interesting!

I'm not familiar with the witness election formula. If somebody only votes for 3 witnesses, is their full voting power spread between those 3, or do they forfeit the rest of their voting power?

I think people should be able to distribute their voting influence to between 1 and 30 witnesses, but they shouldn't be forced to take the cognitive burden of evaluating 30 to maximise their influence if they don't want to.

It reminds of of the famous 'the price of freedom is eternal vigilance'. We have limits to our vigilance. Maybe some can evaluate 30 people (or more), but if the economic incentives are wrong, most will properly evaluate just a few, and just use the rest of their votes because they can, and this is not informed or helpful voting.

@clayop님 어제 파워임대 관련해서 스팀챗으로 연락드렸는데 확인 부탁드립니다!