The outbreak of the COVID-19 pandemic in 2019 isolated the previously connected world. People reduced unnecessary travel and chose to work from home. The COVID-19 pandemic seems to be a public test of social connectivity, in which the necessity and value of connections between people are re-evaluated. People gradually realize that bar gatherings, movie theaters, KTV and other activities that have become an important part of life do not need to exist. The connectivity of the Internet makes up for the isolation of physical space, and social platforms such as WeChat have become the main carriers for people to connect and entertain. #原创 #香港web3嘉年华 #crypto2023

With the booming development of big data and algorithms, online social networking has increasingly become a product of the integration of humans and machines. Social platforms such as WeChat and Weibo not only provide people with a cloud place for social interaction, but also shape people's social methods, thinking habits, and even redefine the friendship between people in all aspects. This article calls the reaction of social media to people the engineering sociality of social media. The engineering sociality of social media has brought many negative effects, such as the negative impact and wrong shaping of social media on human thinking, cognition and behavior. In recent years, there has been more and more discussion on the negative impact of Web2.0 platforms, and many regions and countries have taken actions to restrict social media, such as the recent US proposal to ban Tiktok. In contrast, most of the discussions on Web3.0 are still limited to anti-censorship, ownership, creator economy and other old-fashioned issues that cannot resonate with the public. Therefore, the author would like to explore the negative impact of traditional social media and its enlightenment to Web3.0 from different perspectives, and discuss it in combination with actual projects.

Web3.0 social projects are exploring different paths, such as encrypting communications, introducing ZK technology to protect user privacy, and the data sovereignty movement that decouples data from platforms. What I am most interested in and want to focus on in this article is the social graph. There have been many discussions on social graphs on the Internet, and the mainstream perspective focuses on how social graphs empower developers and improve user experience, but there is not much discussion on the engineering sociality of social graphs. Therefore, the author will take this as the center of this article, and combine the three well-developed projects, CyberConnect, Lens and Farcaster (Warpcast), to analyze the significance of the existence of social graphs and the challenges they face, hoping to trigger some thinking among readers.

Social Graph

The social graph brings together interpersonal relationships on social platforms. Today, the most common relationship on the Internet is "friends" who follow each other. The "friends" here are no longer the same as the original meaning of friends. Social media has extended the meaning of the word "friends". The most primitive human social interaction is limited to a small circle around us due to geographical and temporal limitations. The relationship we establish with the people around us is a strong relationship, and this strong relationship structure is very tight. For example, when we communicate with friends for many years, we don’t need friends to provide me with interesting topics on a regular basis. It is more of a point-to-point information exchange. The "friends" for many years here are fundamentally different from the "friends" created by social media. The relationship brought together by algorithm recommendations is very weak, so "content" is needed to strengthen this relationship. Therefore, when friends with weak relationships socialize, the meaning of communication is reduced, and it is more about content consumption.

In Web2.0, the division of relationships has been reflected in major social software. WeChat accumulates strong relationships, while other social media such as Weibo, Douban, Momo, etc. accumulate weak relationships. In fact, WeChat is no longer a social platform, but more like an address book. The first thing users do after adding friends is to chat, without any content production or consumption. However, on other social software, people will definitely upload their avatars, fill in their information, post a few dynamics, create content, and then establish connections with other people. The main purpose of doing this is to reduce the trust cost between people. Because no one will be willing to socialize with a stranger who has no avatar, blank information, and no dynamics.

From strong relationships to weak relationships, the motivation for content consumption gradually weakens. For example, we like boring daily posts sent by friends on WeChat Moments, but few people are interested in strangers' chatter. In order to make up for the instability of weak relationships and the lack of consumption motivation, social media generally adopts two paths. The first is to rely on high-quality content, and the second is to enhance the connectivity brought by algorithms (discussed in the next section). However, the development of social media that chooses two different paths is diametrically opposed. BBSs that rely on high-quality content and community operations, such as Tieba, Tianya, Douban, etc., have become "tears of the times." And SNS platforms such as Facebook, Twitter, Instagram, etc. dominate the social media rankings. YouTube, which started with community videos, also quickly diluted the concept of community in the middle and late stages, and rapidly expanded using algorithms and recommendation mechanisms to gain a foothold in social media.

Why is it that the more a social platform relies on high-quality content, the lower the value it gets? First, social media needs to mine value from user data. The more connections people make, the more economic value the platform can generate. Therefore, community or small circle culture is not the most conducive social form for platform monetization. Secondly, the more users rely on content, the higher the requirements for the platform's content discovery mechanism. In the era of big data, the accurate transmission of content is an extremely costly thing. Therefore, the algorithm will be inclined to viral eye-catching content instead of pushing high-quality content. Finally, content consumption will eventually flow to strong relationship networks. For example, when we see an interesting content on Weibo, the general operation is to share it with friends on WeChat for content consumption, resulting in the value produced by content-dependent platforms being consumed on other platforms. Or new friends met on Weibo will also add WeChat after getting familiar with each other, and settle on strong relationship network platforms. Therefore, weak-relationship social platforms tend to ignore high-quality content and people's real social experience.

So what enlightenment does the above Web2.0 social media phenomenon have for Web3.0? First, there are differences in "friendships" in different scenarios. The formation of relationships is rooted in the scenario. Secondly, the content distribution mechanism, that is, the algorithm, should be innovated. Next, the author will discuss these two aspects and compare the different paths of the new generation of decentralized social protocols in these two directions.

Social graph scenario

As mentioned above, social graphs are rooted in scenarios. People’s friends on Momo and friends on DingTalk are probably not of the same nature. If the social graph does not distinguish all “connections” by scenarios in the future, it will be extremely difficult to migrate social relationship networks. There is no shortage of examples to prove this. Tencent wanted to establish Tencent Weibo with the users accumulated by QQ Space. The dynamics posted by users on QQ Space will be automatically synchronized to Tencent Weibo. But what Tencent did not consider is that QQ Space is a deposit of acquaintances’ social relationships. It would not be embarrassing to show the “black history” of netizens to familiar people such as family and friends, but if it is pushed to strangers on Weibo, it can be described as a “large-scale social death scene”. The final result can be imagined. Tencent Weibo was defeated by Sina Weibo.

Therefore, social graphs need to be scenario-based. If you want to empower developers, it is far from enough to just provide a wallet follow list. This requires the data to be smaller in granularity and contain richer information. CyberConnect, Lens, and Farcaster have different solutions to this problem. CyberConnect will not limit the scenario to traditional social media, but will embrace the "social +" model at the same time, hoping to integrate social graphs into applications in various fields, such as DeFi, GameFi, Credit, catering, music creation, etc. Therefore, CyberConnect cooperates with third-party projects instead of relying entirely on its own incubation of ecological projects. At the same time, CyberConnect also brings the social assets accumulated in the Web2.0 scenario into Web3.0, connecting the two scenarios of Web2.0 and Web3.0 through Link 3. Therefore, in terms of the depth and breadth of the data, CyberConnect is the better performer among the three.

Lens's scenario-based design is based on content, because Lens modularizes follow relationships and content into NFTs and stores them on the chain. Therefore, people's relationships are not separated from content. Through the content posted, it can be inferred in what scenario a person followed another person. Modularized content and relationships are more convenient for establishing scenarios. In addition, Lens mainly focuses on the social field, and various ecological projects built on Lens are mostly social-related. Because Farcaster has very specific scenarios (Twitter-like applications), the richness and universality of the social graph generated on this platform are also limited. The author believes that this is a big problem in the Farcaster ecosystem.

Social graph-based algorithms

Algorithms are the most important component for facilitating connectivity, and connectivity is the cornerstone of the booming Web2.0 social media, which can help social media maximize network effects. Algorithms change us silently. On social platforms, user autonomy has become an extremely complex concept. Autonomy includes conscious human activities and "technical unconsciousness". To what extent are the social relationships we create on social platforms based on conscious human activities, and to what extent are the connections created by algorithms subtly due to people's "technical unconsciousness"? This question is already difficult to answer today. Because social media will promote "technical unconsciousness" as much as possible, they will first distort the concept of "sharing", equating "invasion of user privacy" with "open and transparent world", and then increase the time users stay on social platforms through a series of coded behaviors, collect a large amount of user data, and finally "cater to their preferences" to guide users from social networks to commercial activities.

For example, Mark Zuckerberg's promise to "make the Internet more social" and his self-proclaimed desire to "make the world more transparent" subtly blur the line between an open Internet and user privacy. NetFlix previously launched a documentary called "Surveillance Capitalism: The Smart Trap". The documentary invited executives from companies such as Google, Facebook, and Twitter to dismantle a series of "addictive" designs built using Internet technology, including: content recommendations, likes, "typing..." and other operations. The only purpose behind this series of designs is to increase the time users stay on the platform and collect as much user behavior as possible. And behind user behavior are similar social norms and cultural logic. For example, the algorithm behind "likes" measures people's desire for things, or their recognition of certain ideas. And this quantified desire can drive potential consumption trends. At the same time, the process of promoting consumption is very invisible. For example, when a user enters Douyin from a link shared by a friend, clicks on the product link at the bottom of the screen, and purchases the product through Alipay, it only takes three clicks to direct the sharing behavior to consumption.

It can be seen that the influence of algorithms on people is subtle and difficult for users to detect. Since earning attention is the first priority of the algorithm, whether high-quality content is distributed is insignificant. The algorithm will tilt traffic to viral eye-catching content. Through these eye-catching fragmented content, users are kept on the platform as long as possible, thereby squeezing attention (such as TikTok). In addition, the personalized recommendations and customization of algorithms may cause people to fall into an information "filter bubble", only receiving information that is consistent with their existing positions, lacking the stimulation and challenge of different opinions, resulting in cognitive bias, information anxiety and blind conformity (information cocoon effect). Social media in the Web2.0 era used algorithms to achieve rapid expansion, but ignored the negative impact of algorithms on people.

In Web3.0, in addition to the recommendation of long-tail content, algorithms based on social graphs should be plural. Vitalik proposed the concept of Plural Intelligence in the article Decentralized Society. Compared with artificial intelligence, the algorithm mechanism under plural intelligence has several major improvements. First, data collection should be rooted in the social context, rather than based on the behavioral characteristics of users on a certain platform; second, data creators, that is, users, should retain the right to govern their data, which is to some extent a confrontation with "technical unconsciousness". In other words, plural algorithms do not make algorithms smarter, but make them more humanistic. Social graphs actually provide soil for plural algorithms. With rich identity information, algorithms can track various characteristics and social backgrounds of users, rather than analyzing based on specific behaviors on a platform. At the same time, if users choose to disclose or hide certain identity information or interpersonal relationships, the model cannot use these data points to customize the algorithm.

From an algorithmic perspective, it is difficult to fundamentally solve the above problems with social graphs alone, because the root of the problem lies in the economic model of Web2.0 social websites, advertising revenue, or in essence, the attention economy. Therefore, Web3.0 social platforms need to use tokens and other media to explore more diverse monetization methods to fundamentally reverse this situation. Social graphs may be able to improve this situation from other levels, such as the accuracy of algorithmic long-tail content push and the control of users over the algorithm.

CyberConnect's infrastructure includes the construction of an algorithm engine. Because the database contains information about user behavior in different applications and scenarios, this engine is of a higher dimension. For example, when building a recommendation engine for a social project, the user's credit on the DeFi platform, the performance of the game platform, etc. can also be analyzed in the algorithm, which is difficult to achieve in the closed context of Web2.0. Lens Protocol currently has no design for algorithms, but it also provides an API, and developers can train their own models through the database. Warpcast, launched by Farcaster, has a recommendation mechanism as a specific product, but this recommendation mechanism is only based on user behavior on its own products. Therefore, although Warpcast has an interface that directly interacts with users and can be used as a tool for customer acquisition and user growth, its flexibility and imagination are also limited because of its too specific product form.