Partnered With:

Vitalik et al. Recent Paper: Decentralized Society -- Searching for the Soul of Web3 (Middle)

Read this article in 40 Minutes

This article talks about how to achieve a richer and more diverse ecosystem through soulbound tokens, as well as key disaggregated property rights and enhanced governance mechanisms in a decentralized society

原文标题：《 Decentralized Society: Finding Web3』s Soul 》

原文作者： E. Glen Weyl, Puja Ohlhaver, Vitalik Buterin

原文编译： Shaun@DAOrayaki.org， DAOrayaki

"Tao is the essence of all things, the treasure of good people, and the protection of bad people."

Chapter 62 Lao Zi

「去中心化社会：找寻 Web3 的灵魂」，是 Vitalik 等人的最新论文，本文讲述了如何通过灵魂绑定 Token 实现更丰富、更多元的生态系统即「去中心化社会（DeSoc）以及去中心化社会中的关键的可分解的产权和增强的治理机制」。因此，DAOrayaki 社区特翻译此文并组织多场 Podcast 深度分析。因全文过长，我们将会分上、中、下三篇发出。本文为该系列中篇，上篇请查看：Vitalik et al. : Searching for the Soul of Web3

4.8 From private and public goods to composite network goods

More broadly, SBTs allows us to effectively represent and manage any assets and goods that lie between fully private and fully public. In reality, even goods consumed by individuals have positive spillovers, just as enabling consumers to contribute better to their families or communities, even the most globally available public goods (climate, for example) are inevitably more useful to some people than others (Seychelles versus Siberia, for example). Similarly, human motivations are rarely entirely selfish or altruistic, and there are many preexisting modes of cooperation, just fewer in some communities and more in others.

However, today's mechanism design assumes that atomized, selfish agents have no pre-existing cooperation, which often leaves mechanisms vulnerable to innocent overcoordination [8]. The worst-case scenario is deliberate collusion between groups that already cooperate. As a result, even the best public financing models, including quadratic financing (QF), cannot scale. QF encourages coordination by offering reduced rewards for collective action by the few, while increasing rewards for collective action by the many.

Ten people split $1 and get a $99 match, resulting in a total of $100, while one person's $10 contribution is not matched. Mathematically, this is done by matching funds to the square of the sum of the square roots of individual donations (we elaborate further in the appendix). But even weak partnerships between large groups like (e.g., most citizens of China) (e.g., donating $1 to a cause) dominate the system and absorb all of its matching funds because QF places a premium on the number of unique contributors. As at present, the QF does not compromise the coordination of relevant special interests, which not only makes the QF mechanism work improperly, but also rewards the relevant special interests.

But rather than seeing pre-existing cooperation as a mistake that we should "rewrite," it's crucial to acknowledge that it actually reflects part of the cooperation that we should leverage and compensate for. After all, we are in the business of encouraging collaboration. The trick is to make quadratic mechanisms work with pre-existing cooperative networks, correcting their bias and tendency to overcoordinate. SBTs provide a natural way for us to tip the scales in favor of collaboration across differences. As Nobel Laureate Eleanor Ostrom has emphasized, the issue is not coordination of public goods per se, but how to help communities of partially cooperative but socially connected individuals overcome their social differences and coordinate at scale across a broader network.

If SBTs represent community membership that reflects the bias of Souls/ Soul clusters, then favoring cooperation across differences simply means discounting the rewards of cooperation for similar or related Souls/ Soul clusters, with similarity measured by their common SBTs. The assumption is that consensus among different Allies is better at creating compound goods that can be applied to a wider network, while consensus among similar Allies is more likely to serve only narrower interests that are over-coordinated (or colluded).

By revealing the memberships between different Souls/ soul sets, SBTs allows us to discount pre-existing collaborations and expand them in quadratic form within emerging networks, giving complex objects to a wider group of interests with the consent of diverse members, rather than giving objects narrow meaning, With the consent of special interests that are innocently over-coordinated (or deliberately colluded). The exact formula for the relevant discount "best" depends on model details and has not yet been studied, but we provide experimental first-hand information for further research in the appendix.

Chapter 5 construction of compound meaning

An example of the variety of online goods that are increasingly prominent in the digital world is predictive models built on user data. Both artificial intelligence (AI) and prediction markets attempt to predict future events based on data obtained primarily from users. But both paradigms are limited in different and almost opposite ways. The dominant paradigm in AI eschewed incentives, collecting data feeds (public or private monitoring) and synthesizing them into predictions through proprietary large-scale non-linear models, in which "USUS" was monopolized using the default Web2, with no "fructus" attributed to data workers.

Prediction markets take the opposite approach, with people betting on outcomes in the hope of achieving financial gains, relying entirely on the economic incentives of financial speculation ("fructus") without synthesising the punters' beliefs to produce combinable models. At the same time, both paradigms produce conclusions that are described as "objective" truths. Ai models are described as "universal" or "universal intelligence," while predictive markets are described as summing up all the beliefs of market participants into a single number: the equilibrium price.

A more productive paradigm is to eschew these extremes and absorb the strengths of both while compensating for their weaknesses and making them more extensive. We propose to combine the complexity of nonlinear AI models with the market incentives of predictive markets to transform passive data workers into active data creators. With this wealth of information rooted in the social nature of data creators, DeSoc is able to unleash a complex intelligent network more powerful than any other approach.

5.1 From forecast market to compound forecast

Prediction markets aim to aggregate beliefs based on wealth and risk appetite from those willing to bet. But this "survival of the fittest" is not an ideal way to aggregate beliefs. A zero-sum game, in which one trader's gain is another trader's loss, assumes that a general predictive power wrestles with "smart people" rather than "stupid people." While wealth may be a proxy for certain abilities and expertise, other forms of relevant expertise may be more reliable predictors. Participants who lose bets in one area may have more accurate beliefs about another. But prediction markets have the unfortunate effect of creating beliefs in those prone to gambling, enriching those who win bets and impoverishing others, and discouraging widespread participation by risk-averse people.

There are better ways to inspire faith. Research shows that while prediction markets generally outperform simple surveys, they do not outperform complex team prediction surveys, which give people an incentive to share and discuss information. Under the team review model, members can weigh in on factors such as past performance and peer review, and teams engage in semi-structured discussions to bring together information that cannot simply be encapsulated in a sale and purchase agreement. Such a team negotiation model can be further refined by using quadratic rules to obtain accurate probability estimates from all participants (compared to prediction markets, which can only obtain an up-down view of the current price balance) [9]. It has been shown that the number of contracts people are motivated to buy reflects their subjective assessment of probability. [10] Such markets would also distribute the benefits of participation more equally, rewarding the right people without bankrupting others, and making everyone a participant in future rounds.

SBTs can unlock a new class of rich models and experiment with predictive power and relative expertise. Prediction markets produce only one number, the price of a contract, and quadratic voting gives each participant's exact belief about the probability of an event. SBTs are able to further calculate these beliefs in the social context of participants' educational credentials, membership, and general sociality to develop better weighted (or non-linear synthetic) prediction models, which are likely to emerge a new generation of expert forecasters at new and unforeseen intersections.

So even if polls don't aggregate beliefs very well, polls can be retrospectively studied to reveal the characteristics of "more correct" participants, and more targeted "experts" can be assembled for future polls, perhaps in the context of a deliberative team. These mechanisms are closely related to those we advocate in this article. Quadratic mechanisms, converted by relevant fractions, can transform poorly coordinated top-down public goods into powerful, bottom-up composite network goods. Similarly, they can transform governance systems based on zero-sum predictive markets into more positive-sum decisions, encouraging the disclosure and synthesis of new and better information.

5.2 From artificial intelligence to compound intelligence

Large-scale nonlinear "neural network" models (e.g. BERT and GPT-3) can also be transformed by SBTs. Such models leverage data from large amounts of public or private monitoring to generate rich models and predictions, such as code based on natural language cues. Most monitored data creators are unaware of their role in creating these models, do not reserve any residual rights for themselves, and are treated as "incidental" rather than key participants.

Furthermore, data collection takes models out of their social context, which masks their biases and limitations and impairs our ability to compensate for them. As more and more high to the requirement of data availability, these contradictions increasingly stand out, new initiatives, such as data source "data collection table", and machine learning privacy laws, these methods need to those who produce data for meaningful economic and management benefit, and inspire them to cooperate to produce separate build model more powerful than them.

SBTs provide a natural way to create economic incentive plans for rich sources of data, while giving data creators residual governance rights over their data. In particular, SBTs allow for careful and targeted incentives for the data (and data quality) of individuals and communities based on their characteristics. At the same time, modelers can track the characteristics and social context of the data collected, as reflected in the SBT, helping to identify contributors who can offset biases and compensate for limitations.

SBTs can also provide customized management rights for data creators, allowing them to form cooperatives, pool data and negotiate usage. This bottom-up programmability of data creators makes possible the future of compound intelligence, where model makers compete to negotiate how to use the same data to build different models. As a result, we move away from an independent, single "AI" paradigm that is separate from our human origins, centralizing surveillance data without a source, and moving to a collaboratively constructed composite

Cambrian intelligences, which are rooted in society and governed by Souls/ soul clusters.

Over time, just as SBT individualizes Soul/ Soul clusters, they also individualize models. Embed data sources, governance, and economic rights directly in the code of the model. Thus, multi-agents, just like humans, build a soul embedded in human sociality, and over time, humans embed themselves in multi-agents, each with a unique soul that complements and cooperates with other souls. Moreover, at this point, we see the convergence of prediction markets and artificial intelligence paradigms in the direction of compound meaning construction. A combination of widely distributed incentives and careful tracking of social contexts creates diverse models that combine the best of both approaches into a technology paradigm more powerful than either.

5.3 Programmable compound privacy

Compound agents raise important questions about data privacy. After all, building such powerful agents requires aggregating data across individuals from large data sets, such as health data, or capturing data that is not interpersonal but shared, such as social graphs.

Advocates of "autonomous identity" tend to view data as private property: because the data of this interaction is mine, I should be able to choose when and to whom I disclose it. In terms of simple private property, however, the data economy is even less well understood than the real economy. In simple two-way relationships, such as an affair, the right to disclose information is usually symmetrical and usually requires the consent and consent of both parties. As the academic Helen Nissenbaum has emphasised, the concern is not "privacy" per se, but the lack of a complete understanding of the context in which information is shared. The Cambridge Analytica scandal centered on people disclosing their social graph attributes and friends' information without their consent.

Rather than viewing privacy as transferable property rights, a more promising approach is to view privacy as a programmable, loosely coupled bundle of rights that allows access, change, or profit from information. Under such a paradigm, each SBT (such as an SBT representing credentials or accessing a data store) would ideally also have implicit programmable property rights and access to some of the underlying information that makes up the SBT, such as holders, agreements between them, shared property (such as data), and obligations to third parties.

For example, some issuers will choose to make SBTs fully public, but some SBTs, such as passports or health records, will be private in the sense of self-sovereignty, and Souls/ Soul Clusters carrying SBT have unilateral disclosure rights. Others, such as an SBT reflecting membership in a data cooperative, involve multiple signatures or more complex community voting rights that all or most SBT holders must agree to disclose.

While there are some technical issues, such as (can SBTs be programmed in this way?) , and important issues surrounding incentive compatibility (further explored in Section 7), but we still believe that programmable compound privacy deserves further investigation and offers key advantages in replacing existing paradigms. With our approach, SBTs has the potential to make privacy a programmable, composable right that maps onto the complex set of expectations and protocols we have today. Furthermore, this programmability can help us rebuild new configurations, as there are countless ways to combine privacy as a right to access information with "USUS", "abusus "and "fructus" to create a subtle cluster of access rights.

For example, SBTs can use specific privacy protection techniques to allow computations against data stores that may be owned and managed by multiple Souls/ Soul clusters. Some SBTS may even allow access to the data in ways that perform certain computations, but the results cannot be proven to third parties. A simple example is voting: the voting mechanism needs to count the votes for each Souls/ Soul cluster, but the votes should not be proven to anyone else to prevent the purchase of votes.

Communication is perhaps the most typical form of sharing data. However, today's communication channels lack both user control and management (" usUS "and" Abusus ") and auction users' attention ("fructus") to the highest bidder, even a robot. SBTs have the potential to manage a healthier "attention economy", giving Souls/ Soul clusters the ability to filter spam from outside their social graph, and possibly even from robots, while improving communication from real communities and desired intersections. Listeners can know more clearly whose voices they are listening to and are better able to assign credit to works that inspire insight.

This economic model is not optimized to maximize engagement, but rather to create a more valuable shared goal through positive sum collaboration. This communication channel is also important for security. As mentioned above, "high bandwidth" communication channels are critical to helping communities build a secure foundation.

Chapter 6: Decentralized society

Web3 hopes to transform society broadly, not just the financial system. However, today's social structures -- keywords such as family, church, team, company, civil society, celebrity, democracy -- would mean nothing if indigenous people did not represent the human soul in the virtual world of wider relationships they espoused (often referred to as the "meta-universe"). If Web3 eschages persistent identity, trust and collaboration models, and assignable rights and permissions, we will see all these over-financialization trends of witch attacks, collusion, and fully transferable private property in a limited economy, respectively.

In order to avoid over-financialization and unleash exponential growth at the same time, we propose enhancing and bridging our sociality in virtual and physical reality, empowering Souls/ Soul clusters and Communities/ community clusters with richer coded social and economic relationships. But trust and cooperation alone are not enough. Correcting biases and tendencies toward excessive coordination (or complicity) in the trust network is critical to encouraging more complex and diverse social relationships than ever before. We call it "decentralized society (DeSoc)" : a co-determined society, in which Souls/ soul clusters and Communities clusters gather together from the bottom up as emerging attributes of each other, producing complex networked objects on different scales.

We emphasize that complex web artifacts are a hallmark of DeSoc because the web is the most powerful engine of economic growth, but also the most easily captured by private actors (such as Web2) and powerful governments. Most of the important economic growth comes from the increase in network revenue, meaning that each additional unit of input produces more output. Examples of simple physical networks include roads, power grids, cities, and other forms of infrastructure that rely on labor and other capital inputs.

Examples of powerful digital networks include markets, predictive models, and compound intelligence built on data. In both cases, network economics differs markedly from neoclassical economics, which is concerned with diminishing returns, with diminishing output for every additional unit of input, whereas private property produces the most efficient results. When private property is applied to the situation of increasing returns, it will have the opposite effect, leading to the phenomenon of restricting the development of the network by extracting rents. A road between the two cities could unlock increasing returns from trade gains. But the same private ownership of roads can stifle growth if owners choose to extract rents from trade between the two cities. Public ownership of the web has its dangers, too, from being captured by regulators or underfunded.

Increasing returns are most effective when the web is seen neither as a purely public good nor as a purely private good, but as a partial and composite public good. DeSoc provides a social basis for the decomposition and reconfiguration of rights - rights of use (" USUS "), rights to consume or destroy (" Abusus ") and rights to gain ("fructus") - and enables effective governance mechanisms for these rights to enhance trust and cooperation, while checking for collusion and capture. We have explored several mechanisms in this article, such as community-based SALSA and quadratic financing (and voting) with associated score discounts. This makes compound ownership a third way, avoiding the Charybdis of private rent and the Scylla of public regulation.

In many ways, DeFi today is a network transformed from a private property model of diminishing returns to one of increasing returns. Built on a foundation of mistrust, DeFi is essentially limited to the realm of fully transferable private property (e.g., transferable tokens), which mainly bundles "usus", "busus" and "fructus". DeFi risks, at best, stifling the growth of the web by charging rents, and at worst, leading to dystopian surveillance monopolies dominated by "whales" harvesting and sucking up data in a race to the bottom, much like Web2.

DeSoc transforms DeFi's race to control and speculate about the value of networks into bottom-up coordination to build, participate and govern networks. At a minimum, DeSoc's social base can make DeFi anti-witch (supporting community governance), anti-vampire (internalizing positive externalities to build open source networks) and anti-collusion (keeping the network decentralized). With DeSoc's structural amendments, DeFi can support and expand diverse networks, broadly empowering benefits, as most of the different members agree, rather than further consolidating networks dominated by narrow interests.

DeSoc's greatest strength, however, is the composability of its network. Increasing returns and network growth do not simply avoid the dangers of rent extraction, but also encourage the proliferation and crossover of nested networks. A road may form a network between two cities. But if cut off from the broader collaboration, the two collaborating cities eventually hit a ceiling of diminishing returns -- either because of crowding (roads and housing) or exhaustion (reaching the limits of the population they can serve). Only through technological innovation and increasingly extensive cooperation, even if that cooperation is loose. Value can continue to multiply by cooperating with neighboring networks to obtain new sources of return. Some cooperation will be tangible, gradually expanding physical commerce across space. But more connections will be informational and digital. Over time, we will see new matrices of cooperation between physical and digital networks that rely on and extend the social interconnections they establish.

It is this intersecting, partially nested, and growing network collaboration structure across the digital and physical worlds that DeSoc facilitates.

By forming networks and coordinating, DaSoc emerges at the intersection of politics and markets, reinforcing both with sociality. DeSoc gave JCR Licklider (founder of ARPANET, which created the Internet) a vision of "man-machine symbiosis" in an "interstellar computer network" and significantly increased social dynamism based on trust. DeSoc is not built on DeFi's trust premise, but codes the trust networks that underpinned today's real economy and enables us to leverage them to generate a variety of online goods that resist capture, extraction, or domination. With this enhanced social aspect, Web3 can sidestep short-term over-financialization in favor of unlimited future revenue growth across social distance.

6.1 Souls can go to heaven or hell

While we selectively highlight the potential that we believe is promising to be unlocked by DeSoc, it's even more important to remember that almost any technology with this transformative potential has the potential to be similarly disruptive: flame burning, wheel crushing, TV brainwashing, car pollution, credit card framed debt, and so on. Here, SBT, which can be used to compensate for intra-group dynamics and achieve cooperation across differences, can also be used to automatically red line undesirable social groups, even target them for cyber or physical attacks, enforce restrictive immigration policies, or make predatory loans. In the current web3 ecosystem, such possibilities do not stand out because they are not meaningful concepts on the current basis. The benefits of enabling DeSoc would also enable these hazards to be realized. Just as the downside of having a heart is that it can be broken, the downside of having a soul is that it can go to hell, the downside of having a society is that it is often driven by hatred, prejudice, violence and fear. Man is a great and often tragic experiment.

As we contemplate DeSoc's possible utopias, we should also situate those possibilities in the context of other technology-enabled dystopias. Web2 is an opaque, authoritarian architecture of surveillance and social control that often relies on top-down, artificial bureaucracies to grant identity (" driver's licenses "), whereas DeSoc relies on horizontal (" peer-to-peer ") social proof. DeSoc authorizes Souls to code their own relationships and co-create multiple attributes, while Web2 mediates or monetizes social relationships through opaque algorithms that can cause polarization, fragmentation, and misinformation. DeSoc eschews the top-down, opaque social credit system. Web2 forms the basis for them. DeSoc treats Souls/ soul clusters as agents, while Web2 treats Souls/ soul clusters as objects.

At least in the short term, the risk of using DeFi for social control (without any identity basis) is small. But DeFi has its own dystopia. While DeFi overcomes explicit forms of centralization -- that is, specific actors have a superior level of formal power in a system -- it has no built-in means to overcome implicit centralization through collusion and market forces. Monopolies don't always emerge the way standard Oil did in the past. Collusion can even occur at higher and farther levels of an ecosystem. We can now see this with the rise of institutional asset managers.

Vanguard, BlackRock, State Street, Fidelity, they're the biggest shareholders in all the big banks, airlines, car companies, and other major industries. Due to these asset management companies stakes in all competitors within an industry (for example, the stakes in every major airlines), their motivation is to make their company look like a competitive industry, but its behavior as a monopolist, at the expense of the interests of consumers and the public to realize the profit maximization and the stability of the whole industry. [11]

在 DeFi 中也是如此，同样的 "鲸鱼 "和风险投资人在堆栈的每一级和堆栈内的竞争者之间积累了更大的股份，也许在 Token 治理中投票，或将其委托给同一类代表，他们在整个网络中也有类似的相关性。如果没有任何对抗女巫的社会基础和强制功能去中心化的相关折扣，我们会看到更多由鲸鱼资助的垄断企业，因为垄断企业越来越成为最大的可用投资资本池。随着「金钱阶层」和用户的分化，我们应该会看到 (而且已经看到) 越来越严重的激励失调和租金榨取。如果出现处理私人数据的 DeFi 应用，我们很可能会看到类似的动态，比如应用鼓励 "拥有 "实际上是人际关系的数据（如他们的社交图）的多人之间的竞标战，以构建单独的私人人工智能，与人类竞争，避免了未来竞争多个人工智能，从而增强人类的能力。

DeSoc therefore does not need to be perfect to pass the test of acceptable non-utopia. To be a model worth exploring, it just needs to be better than the alternatives available. While DeSoc may need to guard against dystopian scenarios, Web2 and the existing DeFi are falling into the inevitable dystopian pattern of concentrating power in the hands of elites who determine social outcomes or own most of the wealth. Web2's direction is deterministic authoritarianism, accelerating the ability of top-down surveillance and behavioral manipulation. The direction of DeFi today is nominally anarcho-capitalist, but has fallen into network effects and monopoly pressures that threaten to turn the medium-term path into authoritarianism in the same way.

DeSoc, by contrast, is a kind of random social pluralism, a network of individuals and communities that collectively determine their own futures as emerging attributes to each other. Viewed from Web2, the development of DeSoc can be compared to the rise of participatory government popular in monarchies over the centuries. Participatory government did not necessarily lead to democracy; it also led to the rise of communism and fascism. Similarly, SBTs does not make digital infrastructure inherently democratic, but democratically compatible based on what Souls/ Soul clusters and Communities/ community clusters jointly decide. Opening up the space of possibility is a clear step forward from Web2's authoritarianism and DeFi's anarcho-capitalism.

reference

[8] We say "innocent" because highly cooperative groups naturally seek to advance their interests, which may well be influenced for their collective good.

[9] Under the quadratic rule, team members can purchase a contract that pays X dollars on the condition that the event occurs, but at a cost of (X^2)/2 dollars. For example, if an event occurs, the individual who sets X=0.5 will receive $0.50, paid by the voter, and will pay at least $0.125 anyway.

[10] If a person evaluates probability P, their expected reward is pX and their cost is X^2/2. Derivative with respect to X, the optimal condition is p=X, assuming risk neutrality, which is reasonable for small bets (rewards and costs can be reduced or increased at will, the same argument still holds).