“To change something, build a new model that makes the existing model obsolete.”

The Game Mechanism of Decentralized Social Media

Dec 2, 2022

A new way of organizing people and people is becoming possible. No matter we call it decentralized social media, DAO or DeSoc, there is a common problem behind it: what kind of mechanism can make the community generate value and achieve a virtuous circle? Such a mechanism requires the community to evolve together. How can we describe and discuss it together?

Since decentralized applications or organizations do not rely on authority institutions, they need to use game mechanisms to replace authority institutions. Through games, individuals and organizations participate in self-government in a market-like manner. It is the first time in history that such an efficient and large-scale game can be realized. How to design such a mechanism is still poorly understood.

In the past few years of discussion and practice, many designs have emerged, corresponding to different problems that a community needs to solve. These patterns of design and questions each apply to one aspect of the community dynamic. When we connect them together, we can describe the general structure of a community.

To truly achieve decentralization, community participation is required to reach a consensus on the design of the mechanism and evolve together. Those feasible and atomic patterns are composable product modules, as well as vocabulary and language for common imagination. In this way, the " pattern language " that allows the community to build together is used in architecture, urban planning, computer language design, product design and other fields, and will also help the establishment of decentralized applications or organizations.

Here are some patterns I've been thinking about recently, all game designs that can be applied to social media or creator communities. Most applications won't use all the modes, but will be able to use some. The patterns below are arranged from the least common, but most fundamental, to the most common, and most personal.

Arbitration using Schelling points: A community will occasionally need to enable arbitration mechanisms, such as maintaining community rules. "Schelling point" provides a decentralized and trustless arbitration method.
Content screening with the prediction market: The process of content creation requires efficient and continuous content governance, and the prediction market provides an efficient and relatively reliable mechanism.
Advertisement space NFT is used as a community shared asset : attention comes from the contribution of the entire community, and it should also be shared by the community. If the advertising space exists in the form of NFT as a common asset of the community, it can use radical market design to realize the path of attention.
Token pledge to curb malicious comments: After the community expands, there will be more and more malicious comments. Even if the author can delete or hide the malicious comments, it is not enough to avoid negative effects. A more effective way is to curb malicious comments by staking tokens.
Community curation with small donations: On the one hand, each community can decide what content is important and valuable, and on the other hand, creators can be motivated by providing valuable content.
SBT marks a personal trust relationship : each person can grant others their own trust, and when this trust is made public, it can open up many mechanisms at the community level. SBT provides a way to expose trust relationships on the chain, providing the basis for other trust mechanisms.

Below is a discussion of each mode.

Arbitration using Schelling points

There is often a need for arbitration in the community, such as whether the content is plagiarized or fake news, and whether an account violates community regulations and needs to be punished. The most intuitive way is to create a small group trusted by people through a certain mechanism to implement the power and obligation of arbitration. This mechanism has many problems. On the one hand, any election process for small groups has low efficiency and low participation rate. On the other hand, the arbitration group can never satisfy everyone, so it needs to bear the pressure from other members.

Ex-Reddit CEO Yishan Wong on Why Content Governance Can't Be Governed Through Moderation Bodies

If we can ensure fairness through the game mechanism, we can randomly select members to participate in decision-making, arbitration can be more efficient, and the results can be more easily reached consensus. We can even make the identity of the arbitrator anonymous, so that the decision-making process is not subject to the pressure of public opinion in the community. The "Schelling point" may be able to become the basis of such an arbitration mechanism.

"Schelling points" are those points where people would choose jointly without cooperation. For example, if a group of people try to meet in New York on a certain day, but they cannot determine the specific time and place, most people may meet in Times Square at twelve o'clock at noon; because the former is the middle of the day, and the latter is the core of New York The landmarks are the Schelling points in time and space, respectively.

We can design an arbitration rule based on this: After staking a certain amount of tokens, the arbitration participants choose the answer to a question; if they agree with the answer of the majority, they will be rewarded, and if they disagree with the answer of the majority, they will lose certain tokens. At this time, participants tend to choose the choice that they think other people will also make, that is, the Schelling point of this question.

Any form of common knowledge ( Common Knowledge ) among the participants is a Schelling point; if the answer to a question resembles common knowledge, it is possible to arbitrate in this form . For example, when a community has a clear definition of personal attacks or plagiarism, it can be used to judge whether an article is a personal attack or plagiarism. Some more complex issues, such as fake news, may also be able to use such a mechanism if a community can reach a consensus on the definition.

In practical situations, there will be situations where "the truth is in the hands of the few", and these situations are also opportunities to promote the evolution of consensus. Such arbitration would need to incorporate re-appeal mechanisms, allowing disputed situations to be re-decided in ways that are more amenable to open discussion and often less efficient. Therefore, the Schelling point will not completely replace the rule of man or centralized decision-making, but will make routine decision-making more decentralized.

In addition to making "negative decisions" about penalties, the mechanism of Schelling points may also be used for "positive decisions" about rewards. For example, if a community selects the selected content of the day every day, can we use the Schelling point mechanism to allow community members to vote in turn to select recognized excellent works? If it works, such a selection of content can also be used for paid subscriptions and become income for the community.

Content filtering with prediction markets

The arbitration mechanism is passive and needs to be triggered to proceed, and each arbitration requires a certain period. The screening of content, whether it is recommended or hidden, often requires efficiency and continuity. Once there is a final way to screen content, whether it is the aforementioned Schelling point arbitration mechanism or the traditional centralized arbitration, we can use the prediction market to expand it to make it more decentralized and efficient.

The idea of the prediction market is that when people bet on the outcome of an event that has not yet happened, the price of each outcome can be used as the group's prediction of this outcome. For example, for the prediction of "whether it will rain tomorrow", we can have "rain coins" and "no rain coins", which can be exchanged for 1 yuan when it rains and does not rain tomorrow respectively. If the market price of "Rain Coin" is 0.8 yuan today and that of "No Rain Coin" is 0.2 yuan, it means that the market predicts that the probability of rain tomorrow is 80%. If "I" thinks it will definitely rain tomorrow, I can buy "Rain Coin" and expect to earn the price difference tomorrow, while at the same time pushing up the price of "Rain Coin".

Because of the relevant interests, people will be more cautious and rational when making decisions, so forecasting markets are generally more accurate than questionnaires. Prediction markets are used not only for the prediction of events such as the presidential election, but also for decentralized decision-making within organizations such as companies, and some scholars have even proposed it as part of the government's operating mechanism . More recently people have predicted whether SBF will go to jail and whether Twitter will go down this year . So can we use the prediction market for more efficient content screening ?

If there is a way to arbitrate violating content, we can randomly select content to apply this mechanism; at the same time, anyone can bet on the outcome of arbitration for any content. For example, if "I" sees a piece of content that is obviously fake news, and the current prediction market predicts that it is not 100% sanctioned, I can bet on "sanctioned" tokens. If this content is not selected for arbitration, then the tokens I bet will be returned in full; if this content is subject to arbitration, the result of the arbitration will determine whether "I" will make a profit or lose money.

Once there is a way to filter high-quality content, the prediction market can be used to filter high-quality content. This screening method can be the aforementioned community content selection, or other mechanisms. For example, in Relevant , users with reputation can "select" the content, while ordinary users can pledge the content with the prediction market mechanism; the income of the pledge depends on whether the final content is selected.

In this way, the results of the prediction market can be used directly on the product to rank the content, or decide whether to hide it or not. On the one hand, readers can give feedback on the content, and on the other hand, they can get rewards for screening content for the community.

Advertisement space NFT as a community common asset

Many communities need external income to keep going, and in addition to the aforementioned paid subscriptions to curated content, advertising is a common source of income. When advertising is tied to community tokens, it also helps stabilize token prices. Because attention to the community comes from the joint efforts of members, advertising revenue should also be shared by community members. We can use advertising slots as public NFTs and apply market rules suitable for public property rights for advertising.

Vitalik once proposed to use the " quadratic payment " method to purchase advertisements, that is, people who place advertisements can spend a small amount of money to occupy a small amount of time, but if they continue to buy time, the price will increase in the form of a quadratic power. Such a mechanism can reduce the cost of ordinary people's participation and allow non-advertisers to participate. The problem with this idea is that it is necessary to ensure that each account is an independent individual, otherwise registering a small account can bypass this rule. But in most communities people need a pseudonym or even anonymity, so this kind of thinking is difficult to work.

Another way of thinking is that, similar to the TheSpace.Game experiment , the Harberger tax is used to trade advertising NFT. At this time, the advertiser buys the NFT at the current price, marks the new price, and pays the corresponding tax; the next advertiser needs to buy the NFT at the new price. This mechanism makes the interests of advertisers and the community more consistent, because the price of the NFT of the advertising space is also the benefit of the advertiser.

In real life, advertisers have many choices, and most of them are not familiar with blockchain-related operations, so it is not realistic for advertisers to directly bid for NFT. At this point, we only need to replace the above advertiser with an advertising agency, and the advertising agency will connect with the advertiser in a traditional and convenient way. The advertising agency can be borne by the application developer at the beginning, but because of the Harberger tax mechanism, there is still sufficient market competition for advertising space, and the interests of the community can still be guaranteed.

Curb malicious comments through token pledge and burning

A common problem with public content is malicious comments or replies. This is especially serious when the number of users is large, it is very common among Twitter users in the currency circle, and there are many examples on Matters. A common way to deal with it is that the original author of the content can hide malicious replies, or report malicious replies. But this process is very inefficient and does not increase the cost of malicious replies, nor does it prevent the damage caused by malicious replies.

One solution is that commenters need to pledge a small amount of tokens for their comments first, and the original author of the comments has the right to destroy these tokens. This increases the cost of malicious commenters, while dynamically determining a community's definition and boundaries of "malicious". In theory, the original author can destroy these tokens maliciously, but because these actions are public, this will make the original author lose readers and commenters.

There are many variations of this idea. For example, a user can only have a "comment pledge", and as long as the pledge is still there, he can comment on other people's content; but any original author who has been commented on can choose to destroy this token, and then the user needs to start again Pledge once. In this way, we can achieve the same game result, but under normal circumstances users only need to make one pledge.

Community content curation with small donations

Community curation is very common. Whether it’s Reddit, Hacker News, Discourse, Pikkick, Douban Group or Weibo hot searches, the content is screened out through behaviors such as “likes”. After the popularity of cryptocurrencies and micropayments, we can replace "likes" with small donations. content quality.

In addition to Matters.News, the Bitcoin community Staker News and the social part of Gitcoin are also experimenting with such content curation methods. If applied at scale, this could be a more versatile and accurate way to measure content value than PageRank. However, presenting content directly through donations is easy to be attacked, because users can control community attention by donating to each other through two accounts, and small groups can also crowd out other content by donating to each other frequently. Some calculations, such as pairwise coordination subsidies in quadratic matching contributions, can alleviate this problem, but cannot completely solve it.

One solution is to give different weights to donation and curation, corresponding to a certain "prestige" of each account; trusted people in the community can accumulate this prestige, while fake accounts are difficult to obtain. For example, people can pledge their community tokens to curators they trust, and then use the pledged amount as the reputation and curation weight. This needs to ensure that the pledger, the curator and the author are not the same person; we can destroy the non-compliant pledged tokens through the aforementioned arbitration mechanism to deter cheating, but it is very difficult to prove whether the two accounts are the same person .

Another idea for designing curation weights is that people record their trust relationship with an account in the form of SBT (Soulbound Token, Soulbound Token), and map it into curation behavior through a certain algorithm when presenting content the weight of. For example, people can authenticate another user they have met in real life and issue the corresponding SBT; when curating, consider how many people have authenticated the curator, and how many other people have authenticated these people. SBT also opens up many other possibilities for content curation, which will be discussed later.

Content curation may be the most valuable part of the community game, because it directly determines what kind of content is seen by more people, but it is also the most unknown part. Behind content curation is a general and far-reaching question: how do we value a public good and provide rewards for it? The creation of public content provides public goods for the community, and its value evaluation will correspond to the aspects of the discovery mechanism and the matching donation mechanism. But once rewards are provided, there needs to be a way to prevent cheating, so that the distribution of resources such as attention and money is fair.

SBT tokenizes personal trust relationship

If a system is open and supports anonymity or pseudonymity at the same time, one person can generate multiple accounts and launch sybil attacks, making content curation, arbitration and other mechanisms have a lot of room for cheating. We can't guarantee (and maybe shouldn't limit?) that a person has only one account, but we can achieve decentralized authentication and let everyone trust other accounts, regardless of whether these addresses are an individual or an organization. This trust relationship can be the basis for the security of the entire system.

Existing social networks already have many such trust relationships. I follow a man by trusting him, trusting that his creations are worthy of my attention and time. The significance of donation is not only a reward, but also a kind of trust, trust in the value of this content. There are also many trust relationships that span online and offline. For example, a friend I have met offline will trust that his account is not a robot account.

A trust relationship like this can be recorded on the chain in the form of Soulbound Token (SBT): SBT, like trust, can be granted or revoked unilaterally, but cannot be transferred. These trust relationships occur between arbitrary users and contain a large amount of information of different dimensions. Once the trust relationship is recorded on the chain, we can design many different mechanisms for solving scenarios that require trust.

For example, in the aforementioned community content curation, we can use the overlap degree of SBT to calculate the similarity between two accounts, and reduce the weight of the curation behavior of highly similar addresses. On the one hand, this increases the difficulty of sybil attacks, and on the other hand, it makes content approved by different communities easier to discover. For another example, a close-knit, half-acquainted community can place the entry threshold on SBT, requiring new members to join the community only after meeting a certain number of old members. The verification of SBT can be combined with Zero Knowledge to ensure the existence of trust relationship, but it does not reveal who is at the other end of the trust chain. This mechanism can be very helpful for social movements, because it not only maintains trust among members, but also guarantees privacy and security.

next question

The above models are either still in the conception stage, or only have a small range of experiments. It's still a long way from a social network that runs primarily on smart contracts. However, these models involving various aspects of community dynamics illustrate that a decentralized social network is possible, and we hope to obtain a better organizational form by replacing the center with games.

To make such an organizational form a reality, it is necessary to distinguish which of these patterns are reliable and can be directly used as the underlying modules, and which are the top-level mechanisms that require continuous discussion and evolution. At the same time, it is also necessary to identify which patterns are suitable as the first application logic to be implemented. "Content curation with small donations" has a natural need in Matters.News, GitCoin, Staker News. Then other mechanisms, such as "token pledge to curb malicious comments", or "use prediction market for content screening", can it make the user experience better in the application and provide value for community members?

Like all other organizations, the sustainability of a decentralized community depends on having a sustainable source of income. Advertising is currently the main source of income for the content industry, but there are many contradictions between the internal motivation of the attention economy and the production of excellent content; direct payment can bring about a healthier content industry, but it is not applicable to public content. For most creators The bar is too high. For a decentralized community, there are more internal economic flow channels, but what kind of mechanism can make external direct payments easier, whether it is donations, paid subscriptions, or something in between? ?