Thesis 3: Reinventing how people connect

With the mass adoption of social graphs, people have been able to connect with other people they knew much more easily. We all have been recommended people on Facebook that are friends of a friend or people that we had met before but were not connected with digitally.

The advent of interest graphs brought further connectivity between people with algorithms that learned from data on what people like - this could be content, merchandise, etc. This enabled people to build communities around their common interests, rather than just with people that they know. TikTok, for example, is currently helping 5 million businesses in the US by helping them build a fan base around their service or product.

One key differentiator between an interest graph like TikTok's and a social graph like Facebook's is the sophistication of the data algorithm and the amount of data tagging it collects on an item that a user likes. Simply put, when the algorithm detects a common item interest from different users, it recommends more items liked by one of the users to the other users based on the common interest tag. On TikTok, this item could be a video, the merchandise in that video, or the topic of the content video content.

However, the interest graph model like TikTok's is ultimately limited by the catalog of the content on the platform. Behavioral data are missing to help us understand if a user performs further actions in pursuit of an interest - making a purchase, joining a community, connecting with other users, etc. These data ultimately would form a behavioral graph that would make recommendations more relevant and actionable for people.

Blockchain as a behavioral graph database

Blockchain as an immutable ledger stores actions as transactions, and by extension, PKA (public key accounts), a.k.a. wallet, helps people access the data stored in that ledger. Over the years of development of various blockchains, it has become a database where almost any data (transactions, assets, social interactions, content, etc.) can be stored, offering a publicly accessible data pool that contains interest data, social data, action data, content data, and much more.

We have already seen the use of trading data on blockchain via queries and analytics (e.g. The Graph, Chainalaysis). These trading data are already being used to evaluate market behaviors and predict market movements. There are also social profiles being created on blockchain, including the various name services and more socially oriented profiles such as CyberProfile by Cyberconnect. All these data provide a natural pool for a behavioral graph.

Building a consumer-facing application utilizing a behavioral graph

To say building on top of a behavioral graph is not to disregard the efficacy of an interest graph in matching users with things they could be interested. We have already seen the success of interest-graph-driven applications in the interest-matching market. There are proven use cases such as content recommendations (e.g. TikTok) and retail shopping suggestions (e.g. Amazon & Alibaba). The key characteristics of a behavioral graph, compared to the interest graph, are as follows:

• Much better knowledge about the user side based on historic behavioral data;

• An attestation layer of the interest graph based on behavioral data;

• Broader application use cases based on behavior categories.

Given the virtually unlimited range of data that can be stored in the blockchain ledger, the question for a company that is building an application on a blockchain is very different from that of a traditional internet application business. In order to provide data-driven services for its users, a traditional internet application would often be forced to focus on one niche market where they are best positioned to collect data. A choice to start off with multiple settings could result in reduced data quality for their initial customer group. Simply put, because a traditional internet startup would not have a blockchain database that contains years of existing data a priori, building an application that caters to multiple business settings would mean a lack of data focus which would then affect service quality.

A blockchain-based consumer-facing application, on the other hand, does not have the data scarcity issue. So instead of building a business model around proprietary data (not to say this is unimportant or unfeasible), we can focus more on discovering and building use cases for the data. In fact, there are already many different use cases in Web 3 that can be catalyzed with data:

• Social connections, which largely rely on manual discovery still in Web 3 at this stage;

• Community membership, which often does not have a discovery process and relies on 3rd party traffic (i.e. from Twitter to Discord);

• Reputation & credit;

• DAO governance;

• Social trading;

• Event planning and recommendations;

• Advertising & user acquisition in Web 3; etc.

Without the data scarcity problem, applications can provide services to all these use cases in one place.

That is not to say that applications that just offer one service are a bad strategy in Web 3, but blockchains offer a much better environment to build super-apps in than the siloed Web 2 databases.

Value capture & distribution of behavioral graph economics

The attribution problem of traditional UGC & social platforms is defined by a centralized approach. Companies own the algorithm and the database that powers the platform and capture all the value created from the user data and the UGC on the platform. As technically all user accounts are controlled in the centralized database, revenue attribution to users and creators goes through means like TikTok's creator fund and YouTube's ad revenue program for creators.

Web 3 primitives catalyze revenue and incentives attribution from the bottom up:

• PKA means the account is owned by the user themself;

• Smart contracts allow revenue & incentive distribution to be done with transparency from the getgo, in group and community settings;

• DAO governance can ensure attribution and incentivization can be adjusted or corrected; etc.

An application that powers individual connections, transaction value, community/project memberships, governance, reputation, content creation, etc. can leverage all these new internet primitives to solve the misalignment between proof of work and incentive attribution.

When it comes to social posting, the original creator and the (effective) reposters that help generate views of the post can receive direct revenue shares to incentivize quality creators and effective reposters; When it comes to community NFT collections, people that participated in the design and DAO governance process of the collection can both receive revenue shares. All this can be achieved with a behavioral graph that allows developers to access past behavioral data to attribute rewards upon provenance or retroactively.

With the right and balanced attribution design, users of the application can feel much more rewarded and recognized compared to being passive consumers on an application built on siloed internet databases with a traditional social graph.

On TowneSquare

At TowneSquare, we are building an everything app with economic attribution that sits on top of a behavioral graph based on public blockchains.