The internet has become an integral part of modern life, providing people with a vast array of resources and opportunities to connect, communicate, and access information. People conduct a wide variety of activities online including but not limited to communication, research, shopping, banking and finance, entertainment, education, social networking, and work.

As we integrate aspects of our lives more fully into online experiences, there is an enormous amount of data collected about our behaviors and preferences in these experiences. This data is used to curate more meaningful experiences for us online via platform-specific algorithms that are able to understand and design for us. This data becomes a representation of who we are online — our persona deduced from our online behavior.

Our persona, however, has not been interoperable in web2 due to ad-revenue business models that encourage companies to hoard any valuable data that they are able to collect in order to maintain a competitive advantage. This siloing of data has prohibited the development of a richer user experience across the web via experiences, services, and products that were not able to scale the proverbial data moat.

We at FirstBatch believe in the democratic principles of an open web where users are empowered with a self-sovereign, privacy-preserving, interoperable ID that allows them to use their data to access the best possible experiences and interactions. We believe that AI and blockchain will fundamentally change the way in which data is generated, collected, and distributed and we are preparing users for that future.

Social Algorithms in Web2

• What Data Sources Do They Rely On

• How Do the Algorithms Work

Self-Sovereign Algorithms

• What does it mean to be self-sovereign

Social Algorithms in Web3

• The Good

• The Bad

• The Ugly

Web3 Projects Case Studies

• Socials

• DAOs/Governance

• DeFi

• CeFi

• Gaming

• NFT Marketplaces

AI and ZK unite forces

• Zero-Knowledge Proofs

• Artificial Intelligence

• Interoperable Personalization algorithms

• Privacy-Oriented Persona Detection Through Social Communities

Conclusion

We all use socials, and for the time being, almost all of them happen to be web2 platforms. Behind the addictive nature of these social platforms, there are algorithms that determine what you will or will not see to optimize your experience on that platform. Obviously, they have various objectives for doing that, including higher time-spent and recurring subscriptions, but the need for gathering user data is common across all personalization and recommendation algorithms.

Internal data (extracted from the user’s actions on the platform):

• Likes, reposts, comments, saved posts

• Friends followed accounts and topics

• Time spent in the app/on a post

• and more

External data:

• Cookies

• Dynamic links

• Access to user contacts, gallery, documents, etc.

• Data from 3rd party data providers

• and more

One of the main challenges with social algorithms is that the collection of data about users is often not under their control. When people use social media platforms and other online services, they often agree to terms of service that allow the companies behind these platforms to collect and use their data in various ways. This can include data about their activity on the platform, as well as data about their online behavior more generally.

As a result, the privacy of users on social media platforms is often limited, as they have little control over the data that is collected about them and how it is used. This can be especially concerning when it comes to sensitive personal information, such as data about their interests, relationships, or health.

Internal data on centralized platforms is often stuck within a single platform and is not easily accessible to other platforms or services. This can be a problem for users, as it means that they are unable to take their data with them if they switch to a different platform, or use it in ways that they might find more useful or convenient.

Social platforms use a variety of techniques to make sense of the data they collected, most of which include AI algorithms and interest graphs. Since the objective is mainly showing users more content that they will like and less of the content they won’t, users’ previous interaction data such as likes or watch time from similar posts get different weights to determine their likeliness to engage with the post in question. While some platforms have self-labeling mechanisms that allow users to categorize their posts (like Reddit) or labels provided by the provider itself (like Netflix), Natural Language Processing algorithms are another solution that allows platforms to group similar posts under the same label in a scalable way.

Interest graphs take the methods mentioned above to another level by creating a graph that allows identifying the relationships between users and content. By utilizing this, social platforms are able to suggest a post liked by a user to another user with similar interests, or suggest a similar post after the user has a high engagement with a piece of content.

While it may improve the users’ experience on the platform by suggesting relevant posts, social algorithms almost always violate user privacy by design.

Self-sovereign identity refers to an individual’s ability to control and manage their own digital identity. This includes being able to access, update, and revoke access to personal information without the need for a central authority or intermediaries.

In order to be considered self-sovereign, social algorithms must meet certain criteria (expanding Christopher Allen’s ten principles):

1. Existence: Users must have an independent existence.

2. Control: Users must control their identities. Algorithms should make the user the ultimate authority on their identity. This includes the user having the right to update or hide any part of their identity at any time. The individual must have the ability to revoke access to their personal information at any time.

3. Access: Users must have access to their own data. They should be aware of all data associated with their identity, nothing can be hidden from users themselves. There should be no actor that restricts the user’s access to data associated with their identity.

4. Transparency: Systems and algorithms must be transparent. Algorithms must be transparent in the way they work and get managed.

5. Persistence: Identities must be long-lived. Without violating users’ right to terminate their data, the identities must exist as long as possible (ideally forever)

6. Portability: Information and services about identity must be transportable

7. Interoperability: Identities should be as widely usable as possible.

8. Consent: Users must agree to the use of their identity.

9. Minimization Disclosure of claims must be minimized

10. Protection: The rights of users must be protected

11. Decentralization: The system must be decentralized and not rely on a central authority to function.

Social algorithms in web3 will differ in mission from social algorithms in web2 in 3 key ways. The first is a shift away from ad-revenue models towards more subscription or token-based revenue models for social networks. One example of this would be social tokens, which allow community creators, influencers, and enterprises to monetize their fan base. Fans who buy social tokens are usually given access to exclusive content and products. Another example can be drawn from Lens protocol where users’ actions like following other users or collecting posts are embodied as NFTs that are sold.

Instead of incentivizing time spent on platforms and the proliferation of viral content for ad revenue, web3 social algorithms aim to provide a relevant experience for their users. The main objective will be to cut through the noise, especially with the advent of AI-generated content.

The second is the source of data that the algorithms are collecting. Contrary to web2 where data is siloed and private, all interactions on web3 socials will be represented as a “transaction” on public blockchains. The availability of data in web3 social will allow the community and users to audit to ensure what networks are delivering is what they’ve committed to delivering. That’s very important to maintain a healthy algorithm ecosystem in web3.

The third is privacy and discretion for users in what is being recommended to them. Given the decentralized and public nature of network data in web3, it becomes more difficult for companies to manipulate a user’s actions on the platform based on their behavioral data. New forms of ID are emerging in web3 that also gives users control over when, with whom, and how much of their social graph is shared.

As we move away from ad-revenue model social networks, users will maintain control of when, with whom, and how much of their identity is shared. This shifts the perception of users from an asset to be farmed to someone who is navigating their own web experience and must be served.

As AI technology advances, AI-generated content is becoming more prevalent on social media platforms and other online environments. This trend has the potential to significantly impact the way that social algorithms work, and the user experience on these platforms.

One of the main effects of AI-generated content is that it can significantly increase the volume of content available on social media platforms. This can be both a blessing and a curse, as it provides users with more choices, but it can also make it more difficult for them to find the most relevant and useful content.

Proof of humanity will be an important concept as the use of artificial intelligence (AI) in generating content becomes more widespread. Proof of humanity would involve verifying that the person behind the content is, in fact, a human, rather than an AI. This will be critical for maintaining the integrity and trustworthiness of online content in an age where AI is increasingly being used to generate and distribute information.

Another is that there will be a new premium on human-generated / bespoke content that will be designed to an individual user’s taste and not catch a trend. These new methods of content creation open up new monetization models for creators and a richer, more relevant experience for users.

In addition to adapting social algorithms to handle AI-generated content, it will also be important to ensure that the user experience on social media platforms remains positive and engaging. This may involve prioritizing user-generated content, or developing new ways to surface the most relevant and useful AI-generated content for each individual user.

Overall, the increasing use of AI-generated content on social media platforms will require ongoing evolution and adaptation of social algorithms, as well as a focus on delivering the best possible user experience.

Interoperability between web 2.0 and web 3.0 is important for a number of reasons. One of the main benefits is that it allows users to access and use the vast amounts of data that have been accumulated on web 2.0 platforms, while also taking advantage of the new capabilities offered by web 3.0.

One of the key challenges in achieving interoperability between the two web versions is the fact that web 2.0 platforms have billions of data points that are not easily accessible to web 3.0 systems.

To address this challenge, it is important to create a bridge between web 2.0 and web 3.0, in order to enable the exchange of data and information between the two. This can be achieved through the use of interoperability standards and protocols, which allow different systems and devices to communicate and share information in a seamless and standardized way.

In addition to the benefits for users, interoperability between web 2.0 and web 3.0 can also be beneficial for web 2.0 companies. For example, consider a company like Netflix, which has a vast amount of data about its users and their viewing habits. If Netflix were able to make this data interoperable with other platforms, such as Twitter, it could potentially create new opportunities for users to connect with relevant content and communities, as well as allow for more personalized recommendations and content suggestions.

While decentralized networks have made great improvements in the accessibility and transparency of web data, there are some caveats. Each blockchain network also has its own independent language and unique apparatus for validating information — a tower of babel scenario. In building a recommendation system that demands composability and efficiency, blockchain networks can be limited by the cost and complexity of sharing data across chains.

Blockchain technology is often associated with financial transactions, as it was originally developed as a way to securely and transparently record transactions using a decentralized database. However, the data stored on a blockchain is typically limited to financial transactions and is not sufficient for calculating social alignment on web 3.0 platforms.

To have effective recommendations on web 3.0 platforms, it is important to have a decentralized identity that is integrated with users’ web 2.0 digital footprints. This includes a wide range of data points beyond just financial transactions, such as social media activity, online behavior, and preferences.

With this broader context, it is easier to gauge a user’s interests and preferences accurately and make genuinely personalized and relevant recommendations.

This will enable web 3.0 platforms to make more accurate and useful recommendations to users and to create more personalized and engaging experiences.

Interoperability and Scaling over the Data Moat:

With greater adoption of blockchain primitives into a user’s “web” experience, there is a strong urge to pit web2 and web3 against each other in a zero-sum game. However, just as there is no singularity of experience in web2, we should expect the same with the advent of mass web3 adoption. The reality is that users will continue to have experiences in both ecosystems simultaneously, and most will gain exposure to web3 primitives through web2.

An important factor to consider is the existing decentralized nature of a user’s experience on web platforms. The average social media user engages with approximately 6.6 platforms, each fulfilling a specific need.

The interoperability of interest data between platforms like Twitter, Reddit, and YouTube can benefit both users and content creators by increasing user engagement across platforms. If users can connect the dots of their cross-platform interests, they may be more likely to spend more time on that platform and interact with other users or content creators. Another consideration is the incorporation of interoperable personalization into the booming Social Commerce industry — valued at 724 billion U.S. dollars in 2022 and is forecasted to surpass six trillion U.S. dollars by 2030. By using social media, Social Commerce consumers can make purchases without leaving a platform, enabling a fast and direct customer journey.

During a 2021 study of social media users from 11 countries, 77 percent of respondents said these platforms inspired them to learn about a product or brand. The same percentage of users said that this channel connected them to brands or products they had never interacted with before.

The key to this discovery and engagement is the social algorithms and interest graphs. Interest graphs are a way to represent users’ interests. Instead of using traditional search terms, expressed in queries such as “Harry Potter,” an interest graph would model interests as nodes and the relationship between them (such as “is a fan of”) using edges. With data composability and interoperability, platforms can build more robust recommendation systems for users as they have a wider and more diverse set of data to draw from.

At present, however, most interest graphs are proprietary and not publicly available, which makes it difficult for third parties to build on top of them. A decentralized interest graph, by contrast, can be more flexible and customizable than a centralized platform because developers can build on top of it and create new features and functionality. This can help drive innovation and enable developers to build applications that better meet the specific needs of their users.

Let’s take Amazon marketplace, for instance, an e-commerce platform with hundreds of millions of users and a massive interest graph. If Amazon were to make its personalization and similarity search algorithms decentralized and interoperable to third-party vendors as a service, this would create enormous value for the users of developers and apps looking to build on top of this infrastructure.

The interoperability of data creates a win-win scenario for both parties: data providers can earn for decentralizing their interest graphs, and developers and apps can more easily determine the market fit and consumer preferences for their products or services.

In web3, we have social blockchains such as DeSo that store all of the data on a public blockchain so that anyone can build and scale a social media experience that’s competitive with the existing data. Businesses can create their own social apps and experiences with these tools without relying on the data silos of major networks or paying an arm and a leg for marketing.

Interoperable interest graphs will lay the foundation for a more scalable, user-centric web experience by lowering the barrier to understanding the fragmented experience of modern web users. The advent of web3 adoption will not be the demise of web2 systems, but rather an extension.

To illustrate the applications of interoperability and personalization further, we have outlined case studies across a few web3 verticals after going through hundreds of web3 projects and analyzing how they can benefit from better personalization.

Needless to say, personalization is at the heart of social platforms. Who would spend hours every day looking at irrelevant posts about topics that they couldn’t care less about? This is why we see advanced levels of personalization and curation in almost all of the most popular social platforms today. Another thing that’s common among a vast majority of these platforms is the lack of privacy and interoperability.

High-quality content curation is not as simple as showing posts that are liked by the people that the user follows — posts should be aligned with the user’s interests in order to keep the experience consistent. This is easier to do within a single platform since you have a limited number of content types and interaction options, but to do this across many platforms requires finding a way that brings all of that data into a form that makes any type of content compared with each other.

When every post from tens of social platforms is represented as vectors that contain relevant labels, it enables use cases that transfer some data & insights from one platform to another. For example, a voice-centric social app like Rally or Clubhouse doesn’t have to rely on the category tags they have internally but instead increases their personalization level recommend people voice channels that align with their interests based on the usage of other platforms that are text or images centric such as Twitter or Instagram. But it is important to keep all of this process completely privacy-preserving, otherwise, it can become too messy too quickly with platforms monetizing users’ data without their permission in any possible way.

This is not even limited to the countless social media platforms that millions of people use. What about a talent platform like tr3butor? Imagine being able to get job recommendations not only based on your CV or LinkedIn profile but also considering a 360 view of your interest that is scattered across many other platforms such as youtube, substack, Twitter, and more. This can be even more flexible with a creative community like Doingud, bringing artists together not only based on self-labeled areas but also on interests detected from all the platforms they choose to spend time on.

DAO formation, as its name suggests, is decentralized in nature. DAOs may organize through social media, real-life networks, events, and hackathons; the possibilities are endless. Due to this expanse of possibility for recruitment, DAOs stand to benefit from a recommendation engine that is able to efficiently connect potential members to organizations and causes that they are closely aligned with through a variety of online channels.

Another pain point for DAOs is voting mechanisms and participation. In traditional DAO structures, token ownership is used to determine who gets to vote on proposals. The problem is that voting power can easily be manipulated by VCs, whales, or other private interests who are able to leverage enough financial capital to buy up governance tokens.

ZK ids with persona traits mapped to a universal interest graph could add a social layer to this voting scheme as a healthier voting mechanism. For example, instead of a simple quorum to decide voting outcomes, eligible voters could be determined by a social weight — a score based on an identity’s social/interest alignment with a given proposal or DAO. Voting based on social alignment could also be a potential solution to low voter engagement in DAO structures.

While the UX of DeFi applications tends to be straightforward and transactional, there are two larger trends with users of DeFi that can benefit from a decentralized recommendation engine. The first is social trading where users are able to follow the trading behavior of their peers and expert traders. The content and interactions of social trading are not limited to DeFi applications — they are decentralized and may occur on multiple social networks in web2 or web3.

To organize and expand the possibilities for interaction amongst social traders, a ZK ID may serve as a useful alternative to users who wish to protect personal information in a potentially risky new environment trading digital assets. User and community recommendation tools could help users safely connect with trading communities that are aligned with their interests without having to create numerous accounts.

The second trend is community investment and philanthropy. As DAOs gain traction, they are becoming a viable alternative to token ICOs and VC funding as a way to bootstrap new projects or distribute grant funding. A decentralized algorithm to connect DAOs with money to spend with projects, founders, or communities who are aligned in the mission would be an incredible growth opportunity for web3 and entrepreneurs as a whole.

Similar mechanisms can be used for DeFi users who are interested in philanthropy and want to discover causes or projects that they are aligned with. While DeFi is usually associated with trading, staking, and other financial primitives, the users of DeFi are plugged into many different areas of web3 with lots of social capital that can be unlocked with interoperable social algorithms.

While there are concerns about the future of CeFi applications, they still serve a critical role in web3 — the fiat on/off ramp. CeFi is often one of the first points of contact for users in web3, and as such can set the tone for a user’s experience. CeFi applications differ from DeFi in that there is more marketing, gamification, feeds, and mechanisms to get users to make transactions to generate fees. For new users, all of this new information can be overwhelming in making decisions about where to begin. However, interoperable personalization can create a win-win situation for users and exchanges. Users will benefit from a more streamlined UX in platforms in finding educational materials, social trading, communities, and more. Central exchanges will benefit from the insights gained about the interests and preferences of communities who use their platform, which they can use in turn to improve their onboarding experience.

In gaming projects, one of the most obvious recommendation concepts is recommending people games that they will probably like. While this is something game stores already do using the data from games the user has played before, this experience can be improved drastically using data from users’ activity from other platforms and from their position on the social graph.

For example, if you have been watching a lot of game walkthrough videos and videos about exploring Mars, youtube will recommend videos that are related to one or both of these topics which makes the experience smooth because you don’t have to search for a new video every time you finish one. But if you sign up to a gaming platform today, you will just get the most popular games at the beginning (and similar to the ones you play as you go).

Let’s imagine a different world. With interoperable personalization, you can sign up to a platform like Ruby Play Network for the first time and get Millions on Mars as a recommendation because you have an interest in that type of game and in that specific topic, based on the videos you watch. The best part is, Ruby Play Network does not collect any data about you or your preferences for this as the entire interaction is encrypted with zero-knowledge proofs.

It doesn’t end there. You can see a hoodie that is very similar to your IRL style being recommended to you in Decentraland or get offered a special quest in The Unfettered because of your interest in a specific period of history that was the inspiration for that side story. In-game recommendations can have an even bigger impact on the gamers’ experience by combining interest data from gaming platforms, livestream platforms, social media, and more.

Ever bought an NFT just for the art? For the community? Utility? For most NFT collectors the answer is yes to at least a few of these types of questions that give NFTs a meaning beyond the floor price and the traded volume. So how come we only see collection/NFT lists that include nothing other than financial and transaction data?

Often times NFTs carry much more than the price history data attached to them including the items they have, art styles used, pop culture references, and community values. You may already be expressing what aesthetic criteria or personal values you have on other platforms, but NFT marketplaces such as OpenSea, Magic Eden, or X2Y2 don’t have a way of knowing that. Wouldn’t it be cool to see your favorite music group’s NFT collection as soon as you open Rarible? Or maybe you care deeply about animal rights and there is a very passionate non-profit working on this cause that wants to collect funds through their NFTs, why should you wait for it to appear on your Twitter feed when it could just pop into the LooksRare recommendations?

NFTs have countless ways of offering personalized solutions to users due to their art and community-centric nature despite being known more for the prices paid.

Zero Knowledge Proofs (ZKPs) are cryptographic concepts that allows one party (the prover) to prove to another party (the verifier) that they know a certain piece of information without actually revealing the information itself. In the context of social algorithms and recommendation engines, ZKPs can be used to ensure the privacy and self-sovereignty of users by allowing them to prove that they have certain preferences or characteristics without actually revealing those preferences or characteristics to the algorithm or to any third party.

Here are some ways in which ZKPs and blockchain technology can be used to ensure the privacy and self-sovereignty of social algorithms and recommendation engines:

• ZKPs can be used to ensure that users’ data is kept private and secure, even when it is being used by a recommendation engine or other algorithm. For example, a user may want to prove to a recommendation engine that they have a certain preference or characteristic (e.g., that they prefer action movies) without actually revealing the specific movies they have watched or liked. The recommendation engine can use a ZKP to verify that the user has this preference without actually seeing the user’s data.

• ZKPs can enable users to control which data they share with algorithms and recommendation engines and to revoke access to their data at any time. For example, a user may want to share their data with a recommendation engine in order to receive personalized recommendations, but they may also want to have the ability to stop sharing their data at any time. A ZKP can be used to enable this kind of selective sharing, allowing the user to maintain control over their data and their privacy.

• ZKPs can be used to ensure that algorithms and recommendation engines do not have access to sensitive or personal information about users. For example, a user may not want to reveal their age, gender, or location to a recommendation engine, but they may still want to receive personalized recommendations based on other factors (e.g., their interests or preferences). A ZKP can be used to enable the recommendation engine to make recommendations based on these other factors without actually seeing the user’s sensitive or personal information.

Overall, ZKPs and blockchain technology offers a number of potential benefits for ensuring the privacy and self-sovereignty of social algorithms and recommendation engines. By enabling users to prove certain facts or characteristics without revealing their actual data, and by allowing users to share their data with algorithms and recommendation engines selectively, these technologies can help protect users’ privacy and give them greater control over their data and their online experiences.

Artificial intelligence (AI) can revolutionize how users are matched with relevant content and other users with similar interests across multiple platforms.

By leveraging advanced machine learning techniques, AI can analyze and understand users’ preferences, behaviors, and interests, and use this knowledge to recommend content and connections that are tailored to their specific needs and interests. Here are some ways in which AI can be used to match users with relevant content and other users with similar interests

Artificial intelligence (AI) has the potential to significantly increase the efficiency of personalization by using interoperable data from various platforms such as Netflix, YouTube, Twitter, and Reddit. In the current web 2.0 landscape, the need for interoperable data damages the user experience (UX) as users are unable to transfer their personalized experiences across platforms.

This can be frustrating for users who have spent time and effort curating their preferences and recommendations on one platform, only to have to start from scratch on another.

One solution to this problem is for AI to aggregate cross-platform personalization data and prioritize content for users on different platforms. This would allow users to seamlessly transition between platforms and have a consistent, personalized experience. For example, a user who frequently watches cooking videos on YouTube and follows food-related accounts on Twitter could have their preferences and interests recognized by an AI system. This system could then recommend similar content on Netflix, such as cooking shows or documentaries about food.

Interoperable Personalization algorithms meet with Generative AI:

AI can also use interoperable data to personalize the content that is presented to users on a given platform. For example, an Interoperable Persona could interact with a generative AI to generate content that the user prefers. It could then use the digital footprints of the user to generate relevant content, such as articles, videos, or newsletters, that the user is likely to enjoy. This type of personalized content generation can help users maximize the discovery of new and interesting material on the whole web, without giving any prompts.

ChatGPT becomes a creator more than a curator:

As AI-powered conversational systems continue to evolve, ChatGPT and similar technologies will likely shift from primarily used to search and answer specific questions to more of a recommendation system. This means that ChatGPT and other chatbots will be able to anticipate and suggest content to users based on their digital footprint, rather than waiting for users to input specific prompts or queries. This type of recommendation system would be more convenient for users, as they would not have to actively search for or request information; instead, they would simply receive the most relevant content based on their interests and past behavior.

Privacy-Oriented Persona Detection Through Social Communities:

Privacy-Preserving Persona Detection can target any interest capable of forming a community on social platforms, a completely privacy-preserving infrastructure. An autonomous, continuous segmentation technology can measure the involvement of any node in a graph with a semantic vector.

Communities are a group of interacting people sharing the same interests. These interests can be a sports team or a specific technology, and they can be both something niche or general. This idea is called homophily. The concept of community is crucial for the decentralized web, next-generation personalization & marketing simply because they possess two essential features. Reaching out is nearly zero effort by definition, considering it is a group of people sharing the same interests. There is no demand for personal data because knowing the interests is possible without recognizing individual behaviors. Therefore, finding or building a valuable community for your mission is of utmost importance.

Considering how the current state of web2 and web3 applications is lacking in terms of many areas such as privacy, personalization, and interoperability and the development potential many verticals have as we discussed in detail, there need to be more solutions that bridge the gap between users’ expectations in those areas.

We work on bringing a privacy-preserving personalization alternative to any platform whether they are building on web2, web3, or anything in between. The interoperability of this solution will open the space up to many different possibilities for new kinds of interactions while giving people the power of controlling their own data with an exceptional user experience that does not reset with every new platform.

The internet has become an integral part of modern life, providing people with a vast array of resources and opportunities to connect, communicate, and access information. People conduct a wide variety of activities online including but not limited to communication, research, shopping, banking and finance, entertainment, education, social networking, and work.

As we integrate aspects of our lives more fully into online experiences, there is an enormous amount of data collected about our behaviors and preferences in these experiences. This data is used to curate more meaningful experiences for us online via platform-specific algorithms that are able to understand and design for us. This data becomes a representation of who we are online — our persona deduced from our online behavior.

Our persona, however, has not been interoperable in web2 due to ad-revenue business models that encourage companies to hoard any valuable data that they are able to collect in order to maintain a competitive advantage. This siloing of data has prohibited the development of a richer user experience across the web via experiences, services, and products that were not able to scale the proverbial data moat.

We at FirstBatch believe in the democratic principles of an open web where users are empowered with a self-sovereign, privacy-preserving, interoperable ID that allows them to use their data to access the best possible experiences and interactions. We believe that AI and blockchain will fundamentally change the way in which data is generated, collected, and distributed and we are preparing users for that future.

Social Algorithms in Web2

• What Data Sources Do They Rely On

• How Do the Algorithms Work

Self-Sovereign Algorithms

• What does it mean to be self-sovereign

Social Algorithms in Web3

• The Good

• The Bad

• The Ugly

Web3 Projects Case Studies

• Socials

• DAOs/Governance

• DeFi

• CeFi

• Gaming

• NFT Marketplaces

AI and ZK unite forces

• Zero-Knowledge Proofs

• Artificial Intelligence

• Interoperable Personalization algorithms

• Privacy-Oriented Persona Detection Through Social Communities

Conclusion

We all use socials, and for the time being, almost all of them happen to be web2 platforms. Behind the addictive nature of these social platforms, there are algorithms that determine what you will or will not see to optimize your experience on that platform. Obviously, they have various objectives for doing that, including higher time-spent and recurring subscriptions, but the need for gathering user data is common across all personalization and recommendation algorithms.

Internal data (extracted from the user’s actions on the platform):

• Likes, reposts, comments, saved posts

• Friends followed accounts and topics

• Time spent in the app/on a post

• and more

External data:

• Cookies

• Dynamic links

• Access to user contacts, gallery, documents, etc.

• Data from 3rd party data providers

• and more

One of the main challenges with social algorithms is that the collection of data about users is often not under their control. When people use social media platforms and other online services, they often agree to terms of service that allow the companies behind these platforms to collect and use their data in various ways. This can include data about their activity on the platform, as well as data about their online behavior more generally.

As a result, the privacy of users on social media platforms is often limited, as they have little control over the data that is collected about them and how it is used. This can be especially concerning when it comes to sensitive personal information, such as data about their interests, relationships, or health.

Internal data on centralized platforms is often stuck within a single platform and is not easily accessible to other platforms or services. This can be a problem for users, as it means that they are unable to take their data with them if they switch to a different platform, or use it in ways that they might find more useful or convenient.

Social platforms use a variety of techniques to make sense of the data they collected, most of which include AI algorithms and interest graphs. Since the objective is mainly showing users more content that they will like and less of the content they won’t, users’ previous interaction data such as likes or watch time from similar posts get different weights to determine their likeliness to engage with the post in question. While some platforms have self-labeling mechanisms that allow users to categorize their posts (like Reddit) or labels provided by the provider itself (like Netflix), Natural Language Processing algorithms are another solution that allows platforms to group similar posts under the same label in a scalable way.

Interest graphs take the methods mentioned above to another level by creating a graph that allows identifying the relationships between users and content. By utilizing this, social platforms are able to suggest a post liked by a user to another user with similar interests, or suggest a similar post after the user has a high engagement with a piece of content.

While it may improve the users’ experience on the platform by suggesting relevant posts, social algorithms almost always violate user privacy by design.

Self-sovereign identity refers to an individual’s ability to control and manage their own digital identity. This includes being able to access, update, and revoke access to personal information without the need for a central authority or intermediaries.

In order to be considered self-sovereign, social algorithms must meet certain criteria (expanding Christopher Allen’s ten principles):

1. Existence: Users must have an independent existence.

2. Control: Users must control their identities. Algorithms should make the user the ultimate authority on their identity. This includes the user having the right to update or hide any part of their identity at any time. The individual must have the ability to revoke access to their personal information at any time.

3. Access: Users must have access to their own data. They should be aware of all data associated with their identity, nothing can be hidden from users themselves. There should be no actor that restricts the user’s access to data associated with their identity.

4. Transparency: Systems and algorithms must be transparent. Algorithms must be transparent in the way they work and get managed.

5. Persistence: Identities must be long-lived. Without violating users’ right to terminate their data, the identities must exist as long as possible (ideally forever)

6. Portability: Information and services about identity must be transportable

7. Interoperability: Identities should be as widely usable as possible.

8. Consent: Users must agree to the use of their identity.

9. Minimization Disclosure of claims must be minimized

10. Protection: The rights of users must be protected

11. Decentralization: The system must be decentralized and not rely on a central authority to function.

Social algorithms in web3 will differ in mission from social algorithms in web2 in 3 key ways. The first is a shift away from ad-revenue models towards more subscription or token-based revenue models for social networks. One example of this would be social tokens, which allow community creators, influencers, and enterprises to monetize their fan base. Fans who buy social tokens are usually given access to exclusive content and products. Another example can be drawn from Lens protocol where users’ actions like following other users or collecting posts are embodied as NFTs that are sold.

Instead of incentivizing time spent on platforms and the proliferation of viral content for ad revenue, web3 social algorithms aim to provide a relevant experience for their users. The main objective will be to cut through the noise, especially with the advent of AI-generated content.

The second is the source of data that the algorithms are collecting. Contrary to web2 where data is siloed and private, all interactions on web3 socials will be represented as a “transaction” on public blockchains. The availability of data in web3 social will allow the community and users to audit to ensure what networks are delivering is what they’ve committed to delivering. That’s very important to maintain a healthy algorithm ecosystem in web3.

The third is privacy and discretion for users in what is being recommended to them. Given the decentralized and public nature of network data in web3, it becomes more difficult for companies to manipulate a user’s actions on the platform based on their behavioral data. New forms of ID are emerging in web3 that also gives users control over when, with whom, and how much of their social graph is shared.

As we move away from ad-revenue model social networks, users will maintain control of when, with whom, and how much of their identity is shared. This shifts the perception of users from an asset to be farmed to someone who is navigating their own web experience and must be served.

As AI technology advances, AI-generated content is becoming more prevalent on social media platforms and other online environments. This trend has the potential to significantly impact the way that social algorithms work, and the user experience on these platforms.

One of the main effects of AI-generated content is that it can significantly increase the volume of content available on social media platforms. This can be both a blessing and a curse, as it provides users with more choices, but it can also make it more difficult for them to find the most relevant and useful content.

Proof of humanity will be an important concept as the use of artificial intelligence (AI) in generating content becomes more widespread. Proof of humanity would involve verifying that the person behind the content is, in fact, a human, rather than an AI. This will be critical for maintaining the integrity and trustworthiness of online content in an age where AI is increasingly being used to generate and distribute information.

Another is that there will be a new premium on human-generated / bespoke content that will be designed to an individual user’s taste and not catch a trend. These new methods of content creation open up new monetization models for creators and a richer, more relevant experience for users.

In addition to adapting social algorithms to handle AI-generated content, it will also be important to ensure that the user experience on social media platforms remains positive and engaging. This may involve prioritizing user-generated content, or developing new ways to surface the most relevant and useful AI-generated content for each individual user.

Overall, the increasing use of AI-generated content on social media platforms will require ongoing evolution and adaptation of social algorithms, as well as a focus on delivering the best possible user experience.

Interoperability between web 2.0 and web 3.0 is important for a number of reasons. One of the main benefits is that it allows users to access and use the vast amounts of data that have been accumulated on web 2.0 platforms, while also taking advantage of the new capabilities offered by web 3.0.

One of the key challenges in achieving interoperability between the two web versions is the fact that web 2.0 platforms have billions of data points that are not easily accessible to web 3.0 systems.

To address this challenge, it is important to create a bridge between web 2.0 and web 3.0, in order to enable the exchange of data and information between the two. This can be achieved through the use of interoperability standards and protocols, which allow different systems and devices to communicate and share information in a seamless and standardized way.

In addition to the benefits for users, interoperability between web 2.0 and web 3.0 can also be beneficial for web 2.0 companies. For example, consider a company like Netflix, which has a vast amount of data about its users and their viewing habits. If Netflix were able to make this data interoperable with other platforms, such as Twitter, it could potentially create new opportunities for users to connect with relevant content and communities, as well as allow for more personalized recommendations and content suggestions.

While decentralized networks have made great improvements in the accessibility and transparency of web data, there are some caveats. Each blockchain network also has its own independent language and unique apparatus for validating information — a tower of babel scenario. In building a recommendation system that demands composability and efficiency, blockchain networks can be limited by the cost and complexity of sharing data across chains.

Blockchain technology is often associated with financial transactions, as it was originally developed as a way to securely and transparently record transactions using a decentralized database. However, the data stored on a blockchain is typically limited to financial transactions and is not sufficient for calculating social alignment on web 3.0 platforms.

To have effective recommendations on web 3.0 platforms, it is important to have a decentralized identity that is integrated with users’ web 2.0 digital footprints. This includes a wide range of data points beyond just financial transactions, such as social media activity, online behavior, and preferences.

With this broader context, it is easier to gauge a user’s interests and preferences accurately and make genuinely personalized and relevant recommendations.

This will enable web 3.0 platforms to make more accurate and useful recommendations to users and to create more personalized and engaging experiences.

Interoperability and Scaling over the Data Moat:

With greater adoption of blockchain primitives into a user’s “web” experience, there is a strong urge to pit web2 and web3 against each other in a zero-sum game. However, just as there is no singularity of experience in web2, we should expect the same with the advent of mass web3 adoption. The reality is that users will continue to have experiences in both ecosystems simultaneously, and most will gain exposure to web3 primitives through web2.

An important factor to consider is the existing decentralized nature of a user’s experience on web platforms. The average social media user engages with approximately 6.6 platforms, each fulfilling a specific need.

The interoperability of interest data between platforms like Twitter, Reddit, and YouTube can benefit both users and content creators by increasing user engagement across platforms. If users can connect the dots of their cross-platform interests, they may be more likely to spend more time on that platform and interact with other users or content creators. Another consideration is the incorporation of interoperable personalization into the booming Social Commerce industry — valued at 724 billion U.S. dollars in 2022 and is forecasted to surpass six trillion U.S. dollars by 2030. By using social media, Social Commerce consumers can make purchases without leaving a platform, enabling a fast and direct customer journey.

During a 2021 study of social media users from 11 countries, 77 percent of respondents said these platforms inspired them to learn about a product or brand. The same percentage of users said that this channel connected them to brands or products they had never interacted with before.

The key to this discovery and engagement is the social algorithms and interest graphs. Interest graphs are a way to represent users’ interests. Instead of using traditional search terms, expressed in queries such as “Harry Potter,” an interest graph would model interests as nodes and the relationship between them (such as “is a fan of”) using edges. With data composability and interoperability, platforms can build more robust recommendation systems for users as they have a wider and more diverse set of data to draw from.

At present, however, most interest graphs are proprietary and not publicly available, which makes it difficult for third parties to build on top of them. A decentralized interest graph, by contrast, can be more flexible and customizable than a centralized platform because developers can build on top of it and create new features and functionality. This can help drive innovation and enable developers to build applications that better meet the specific needs of their users.

Let’s take Amazon marketplace, for instance, an e-commerce platform with hundreds of millions of users and a massive interest graph. If Amazon were to make its personalization and similarity search algorithms decentralized and interoperable to third-party vendors as a service, this would create enormous value for the users of developers and apps looking to build on top of this infrastructure.

The interoperability of data creates a win-win scenario for both parties: data providers can earn for decentralizing their interest graphs, and developers and apps can more easily determine the market fit and consumer preferences for their products or services.

In web3, we have social blockchains such as DeSo that store all of the data on a public blockchain so that anyone can build and scale a social media experience that’s competitive with the existing data. Businesses can create their own social apps and experiences with these tools without relying on the data silos of major networks or paying an arm and a leg for marketing.

Interoperable interest graphs will lay the foundation for a more scalable, user-centric web experience by lowering the barrier to understanding the fragmented experience of modern web users. The advent of web3 adoption will not be the demise of web2 systems, but rather an extension.

To illustrate the applications of interoperability and personalization further, we have outlined case studies across a few web3 verticals after going through hundreds of web3 projects and analyzing how they can benefit from better personalization.

Needless to say, personalization is at the heart of social platforms. Who would spend hours every day looking at irrelevant posts about topics that they couldn’t care less about? This is why we see advanced levels of personalization and curation in almost all of the most popular social platforms today. Another thing that’s common among a vast majority of these platforms is the lack of privacy and interoperability.

High-quality content curation is not as simple as showing posts that are liked by the people that the user follows — posts should be aligned with the user’s interests in order to keep the experience consistent. This is easier to do within a single platform since you have a limited number of content types and interaction options, but to do this across many platforms requires finding a way that brings all of that data into a form that makes any type of content compared with each other.

When every post from tens of social platforms is represented as vectors that contain relevant labels, it enables use cases that transfer some data & insights from one platform to another. For example, a voice-centric social app like Rally or Clubhouse doesn’t have to rely on the category tags they have internally but instead increases their personalization level recommend people voice channels that align with their interests based on the usage of other platforms that are text or images centric such as Twitter or Instagram. But it is important to keep all of this process completely privacy-preserving, otherwise, it can become too messy too quickly with platforms monetizing users’ data without their permission in any possible way.

This is not even limited to the countless social media platforms that millions of people use. What about a talent platform like tr3butor? Imagine being able to get job recommendations not only based on your CV or LinkedIn profile but also considering a 360 view of your interest that is scattered across many other platforms such as youtube, substack, Twitter, and more. This can be even more flexible with a creative community like Doingud, bringing artists together not only based on self-labeled areas but also on interests detected from all the platforms they choose to spend time on.

DAO formation, as its name suggests, is decentralized in nature. DAOs may organize through social media, real-life networks, events, and hackathons; the possibilities are endless. Due to this expanse of possibility for recruitment, DAOs stand to benefit from a recommendation engine that is able to efficiently connect potential members to organizations and causes that they are closely aligned with through a variety of online channels.

Another pain point for DAOs is voting mechanisms and participation. In traditional DAO structures, token ownership is used to determine who gets to vote on proposals. The problem is that voting power can easily be manipulated by VCs, whales, or other private interests who are able to leverage enough financial capital to buy up governance tokens.

ZK ids with persona traits mapped to a universal interest graph could add a social layer to this voting scheme as a healthier voting mechanism. For example, instead of a simple quorum to decide voting outcomes, eligible voters could be determined by a social weight — a score based on an identity’s social/interest alignment with a given proposal or DAO. Voting based on social alignment could also be a potential solution to low voter engagement in DAO structures.

While the UX of DeFi applications tends to be straightforward and transactional, there are two larger trends with users of DeFi that can benefit from a decentralized recommendation engine. The first is social trading where users are able to follow the trading behavior of their peers and expert traders. The content and interactions of social trading are not limited to DeFi applications — they are decentralized and may occur on multiple social networks in web2 or web3.

To organize and expand the possibilities for interaction amongst social traders, a ZK ID may serve as a useful alternative to users who wish to protect personal information in a potentially risky new environment trading digital assets. User and community recommendation tools could help users safely connect with trading communities that are aligned with their interests without having to create numerous accounts.

The second trend is community investment and philanthropy. As DAOs gain traction, they are becoming a viable alternative to token ICOs and VC funding as a way to bootstrap new projects or distribute grant funding. A decentralized algorithm to connect DAOs with money to spend with projects, founders, or communities who are aligned in the mission would be an incredible growth opportunity for web3 and entrepreneurs as a whole.

Similar mechanisms can be used for DeFi users who are interested in philanthropy and want to discover causes or projects that they are aligned with. While DeFi is usually associated with trading, staking, and other financial primitives, the users of DeFi are plugged into many different areas of web3 with lots of social capital that can be unlocked with interoperable social algorithms.

While there are concerns about the future of CeFi applications, they still serve a critical role in web3 — the fiat on/off ramp. CeFi is often one of the first points of contact for users in web3, and as such can set the tone for a user’s experience. CeFi applications differ from DeFi in that there is more marketing, gamification, feeds, and mechanisms to get users to make transactions to generate fees. For new users, all of this new information can be overwhelming in making decisions about where to begin. However, interoperable personalization can create a win-win situation for users and exchanges. Users will benefit from a more streamlined UX in platforms in finding educational materials, social trading, communities, and more. Central exchanges will benefit from the insights gained about the interests and preferences of communities who use their platform, which they can use in turn to improve their onboarding experience.

In gaming projects, one of the most obvious recommendation concepts is recommending people games that they will probably like. While this is something game stores already do using the data from games the user has played before, this experience can be improved drastically using data from users’ activity from other platforms and from their position on the social graph.

For example, if you have been watching a lot of game walkthrough videos and videos about exploring Mars, youtube will recommend videos that are related to one or both of these topics which makes the experience smooth because you don’t have to search for a new video every time you finish one. But if you sign up to a gaming platform today, you will just get the most popular games at the beginning (and similar to the ones you play as you go).

Let’s imagine a different world. With interoperable personalization, you can sign up to a platform like Ruby Play Network for the first time and get Millions on Mars as a recommendation because you have an interest in that type of game and in that specific topic, based on the videos you watch. The best part is, Ruby Play Network does not collect any data about you or your preferences for this as the entire interaction is encrypted with zero-knowledge proofs.

It doesn’t end there. You can see a hoodie that is very similar to your IRL style being recommended to you in Decentraland or get offered a special quest in The Unfettered because of your interest in a specific period of history that was the inspiration for that side story. In-game recommendations can have an even bigger impact on the gamers’ experience by combining interest data from gaming platforms, livestream platforms, social media, and more.

Ever bought an NFT just for the art? For the community? Utility? For most NFT collectors the answer is yes to at least a few of these types of questions that give NFTs a meaning beyond the floor price and the traded volume. So how come we only see collection/NFT lists that include nothing other than financial and transaction data?

Often times NFTs carry much more than the price history data attached to them including the items they have, art styles used, pop culture references, and community values. You may already be expressing what aesthetic criteria or personal values you have on other platforms, but NFT marketplaces such as OpenSea, Magic Eden, or X2Y2 don’t have a way of knowing that. Wouldn’t it be cool to see your favorite music group’s NFT collection as soon as you open Rarible? Or maybe you care deeply about animal rights and there is a very passionate non-profit working on this cause that wants to collect funds through their NFTs, why should you wait for it to appear on your Twitter feed when it could just pop into the LooksRare recommendations?

NFTs have countless ways of offering personalized solutions to users due to their art and community-centric nature despite being known more for the prices paid.

Zero Knowledge Proofs (ZKPs) are cryptographic concepts that allows one party (the prover) to prove to another party (the verifier) that they know a certain piece of information without actually revealing the information itself. In the context of social algorithms and recommendation engines, ZKPs can be used to ensure the privacy and self-sovereignty of users by allowing them to prove that they have certain preferences or characteristics without actually revealing those preferences or characteristics to the algorithm or to any third party.

Here are some ways in which ZKPs and blockchain technology can be used to ensure the privacy and self-sovereignty of social algorithms and recommendation engines:

• ZKPs can be used to ensure that users’ data is kept private and secure, even when it is being used by a recommendation engine or other algorithm. For example, a user may want to prove to a recommendation engine that they have a certain preference or characteristic (e.g., that they prefer action movies) without actually revealing the specific movies they have watched or liked. The recommendation engine can use a ZKP to verify that the user has this preference without actually seeing the user’s data.

• ZKPs can enable users to control which data they share with algorithms and recommendation engines and to revoke access to their data at any time. For example, a user may want to share their data with a recommendation engine in order to receive personalized recommendations, but they may also want to have the ability to stop sharing their data at any time. A ZKP can be used to enable this kind of selective sharing, allowing the user to maintain control over their data and their privacy.

• ZKPs can be used to ensure that algorithms and recommendation engines do not have access to sensitive or personal information about users. For example, a user may not want to reveal their age, gender, or location to a recommendation engine, but they may still want to receive personalized recommendations based on other factors (e.g., their interests or preferences). A ZKP can be used to enable the recommendation engine to make recommendations based on these other factors without actually seeing the user’s sensitive or personal information.

Overall, ZKPs and blockchain technology offers a number of potential benefits for ensuring the privacy and self-sovereignty of social algorithms and recommendation engines. By enabling users to prove certain facts or characteristics without revealing their actual data, and by allowing users to share their data with algorithms and recommendation engines selectively, these technologies can help protect users’ privacy and give them greater control over their data and their online experiences.

Artificial intelligence (AI) can revolutionize how users are matched with relevant content and other users with similar interests across multiple platforms.

By leveraging advanced machine learning techniques, AI can analyze and understand users’ preferences, behaviors, and interests, and use this knowledge to recommend content and connections that are tailored to their specific needs and interests. Here are some ways in which AI can be used to match users with relevant content and other users with similar interests

Artificial intelligence (AI) has the potential to significantly increase the efficiency of personalization by using interoperable data from various platforms such as Netflix, YouTube, Twitter, and Reddit. In the current web 2.0 landscape, the need for interoperable data damages the user experience (UX) as users are unable to transfer their personalized experiences across platforms.

This can be frustrating for users who have spent time and effort curating their preferences and recommendations on one platform, only to have to start from scratch on another.

One solution to this problem is for AI to aggregate cross-platform personalization data and prioritize content for users on different platforms. This would allow users to seamlessly transition between platforms and have a consistent, personalized experience. For example, a user who frequently watches cooking videos on YouTube and follows food-related accounts on Twitter could have their preferences and interests recognized by an AI system. This system could then recommend similar content on Netflix, such as cooking shows or documentaries about food.

Interoperable Personalization algorithms meet with Generative AI:

AI can also use interoperable data to personalize the content that is presented to users on a given platform. For example, an Interoperable Persona could interact with a generative AI to generate content that the user prefers. It could then use the digital footprints of the user to generate relevant content, such as articles, videos, or newsletters, that the user is likely to enjoy. This type of personalized content generation can help users maximize the discovery of new and interesting material on the whole web, without giving any prompts.

ChatGPT becomes a creator more than a curator:

As AI-powered conversational systems continue to evolve, ChatGPT and similar technologies will likely shift from primarily used to search and answer specific questions to more of a recommendation system. This means that ChatGPT and other chatbots will be able to anticipate and suggest content to users based on their digital footprint, rather than waiting for users to input specific prompts or queries. This type of recommendation system would be more convenient for users, as they would not have to actively search for or request information; instead, they would simply receive the most relevant content based on their interests and past behavior.

Privacy-Oriented Persona Detection Through Social Communities:

Privacy-Preserving Persona Detection can target any interest capable of forming a community on social platforms, a completely privacy-preserving infrastructure. An autonomous, continuous segmentation technology can measure the involvement of any node in a graph with a semantic vector.

Communities are a group of interacting people sharing the same interests. These interests can be a sports team or a specific technology, and they can be both something niche or general. This idea is called homophily. The concept of community is crucial for the decentralized web, next-generation personalization & marketing simply because they possess two essential features. Reaching out is nearly zero effort by definition, considering it is a group of people sharing the same interests. There is no demand for personal data because knowing the interests is possible without recognizing individual behaviors. Therefore, finding or building a valuable community for your mission is of utmost importance.

Considering how the current state of web2 and web3 applications is lacking in terms of many areas such as privacy, personalization, and interoperability and the development potential many verticals have as we discussed in detail, there need to be more solutions that bridge the gap between users’ expectations in those areas.

We work on bringing a privacy-preserving personalization alternative to any platform whether they are building on web2, web3, or anything in between. The interoperability of this solution will open the space up to many different possibilities for new kinds of interactions while giving people the power of controlling their own data with an exceptional user experience that does not reset with every new platform.