Building ‘Google For The Economic Web’ on The Ethereum Blockchain.

Userfeeds is the content ranking & reputation system for blockchain communities. It is inspired by the ideas of liquid democracy and Google PageRank. In the following article, we introduce its vision, economics and the incentive model.

14 min readNov 2, 2016

For a TL;DR version you can just browse through the slides from the presentation I gave on the same topic. If you prefer to watch the entire presentation you can find it here.

Introduction

In the early 1990s when the Web 1.0 was in its infancy, the dominant method of discovering websites was through online catalogues like Yahoo. We applied the old discovery model of browsing topically organised paper catalogues to the new medium of digital websites.

The business model was based on keeping users on a site, clicking through the catalogue. Paper catalogues edited by people were around for years and websites were digital ‘pages’ after all. So the ‘directory of digital pages’ was the natural answer to the website discovery problem.

When Larry Page and Sergey Brin wanted to sell their Google search engine in 1998 for $1 million, they were rejected. Their big idea expressed in Google’s PageRank algorithm was that the Web required a new discovery model — a search engine based on links between websites rather than top-down directories. At the time it was a contrarian thought but today we know they were right.

The Economic Web is About People

In our view, we’re at the early days of the new paradigm — the Economic Web enabled by the blockchains and decentralised technologies.

Blockchains can be viewed as distributed databases with access control based on public-private key cryptography. You can find countless discussions on the web regarding the definition what blockchains really are. To us, blockchains are what routers, servers, and networking protocols were to the early Web — the computer-to-computer communication layer that enables new way of human interaction on the layers above.

The Economic Web is the web of humans transacting as economic agents using the blockchains as the interaction layer.

Tokens are new links between people

In our model, tokens are a critical component of the emerging Economic Web. Tokens are for people what code is for machines — a means of programming their behaviours.

Bitcoin proved that it’s possible to use the concept of a token to change economic behaviours of a vast number of people, by linking them into a network and making them perform activity for the network. Tokens proved to be extremely effective motivators as they utilize survival instincts of our brains triggering experiences of gain, loss, fear-of-missing-out on opportunity etc.

We are naturally hardwired to seek economic benefits and tokens exploit that quality of our psychology. We program them but then they start programming us. To economically incentivize large groups of people, distributed around the world, just a few years ago you required to be a government or a multinational corporation. Today, every smart 15-year old with a laptop can achieve the same result with the cryptographic token. Our survival instincts don’t distinguish if the incentive is an apple on a tree, a corporate stock or a cryptographic token you can sell. People will do stuff for the reward.

Economic Networks or the Web 3.0 consist of networks of people linked by ownership of same tokens (there are already thousands of them). This evolution is illustrated in this table:

The Economic Web Needs a New Content Discovery Paradigm.

“History doesn’t repeat itself but it often rhymes” — Mark Twain.

As you can see from the table, we don’t have the native content discovery model for Economic Networks.

It’s a similar problem that the early Web had in the 1990s. The new type of paradigm emerges but we apply the old content discovery model to it.

What Yahoo catalogues and directories were to websites, social media and search are to economic networks. Going to FB/Twitter or searching Google about any type of cryptotoken is like going to Yahoo directory to browse the web.

You’re going to get suboptimal results. Why?

Yahoo completely ignored the wealth of reputational signals that were stored in links between websites. They opted for the human curated directory instead. So they missed critical signals that were native to the network they wanted to curate. Google exploited that with PageRank and won.

In the similar fashion, Google/FB/Twitter/Reddit completely ignore reputational signals of the token-based economic networks (Bitcoin, Ethereum and many others) to rank information on their platforms.

Do you want to see the result of that?

Or just visit any token-related subreddit on any day.

Distributed Interest Groups

These autocomplete suggestions Google gives you are the effect of the activity of Distributed Interest Groups. While various interest groups pushing their own agenda and interests are nothing new, they were historically limited to certain geographies with fixed memberships. You could tell who is in, and who is out.

Distributed Interest Groups are a new phenomenon enabled by tokens running on top of public blockchain networks. Tens of thousands of people can belong to them without disclosing that membership to others. You just need to buy a token to join or sell it to leave.

Once people join the network they have a vested interest in pushing the same agenda and increasing chances of success for their token. These activities range from ‘to the moon!’ type of posts on Reddit, to elaborate technical blog posts by experts explaining why token X is the future and token Y is doomed to fail.

Tokens not only make people burn electricity, but they also make people say things online.

But the ranking algorithms are completely clueless regarding economics & incentives behind these statements. Google can’t check how many X tokens you own and count that into ranking of your website. Facebook can suggest you articles liked by your friends but it can’t tell you which ones were endorsed by people owning at least 1% of tokens you own.

This is the problem we want to address with Userfeeds.

userfeeds.io is a content ranking and reputation system inspired by the ideas of liquid democracy and Google’s PageRank algorithm.

It can be used as a foundation for content discovery applications based on economic ranking factors.

Facebook answers the question:

“what/who should I pay attention to based on my social links (same friends, interests)”

Applications based on Userfeeds will answer the question:

“what/who should I pay attention to based on my cryptoeconomic links (same tokens owned, same economic incentives)”

Liquid Democracy Meets Google PageRank

Token Holders Are Content Curators

The core concept of our model is that tokens become reputational signals used for ranking of information.

Token holders become Content Curators and can use their token holdings to:

endorse content on the Web
back other curators

Endorsements and backings are timestamped transactions that transfer reputation instead of monetary value. And while the timestamp is objective the reputation is calculated individually and subjectively by each app.

For example: account #1 owning 1000 ETH backs account #2 owning 1 ETH. No actual ETH changed hands. Both accounts still own same amounts. But account #2 now owns 1001 of ETH of app-level reputation (as long as 1000 ETH is present on account #1, the app monitors the blockchain for that)

You can think of ‘reputation’ transfers as links or likes. They create a permanent relationship on the blockchain between identities and content.

It is a combination of liquid democracy and Google’s PageRank.

The curation graph generated by activities of curators is open, permanent and publicly visible on the blockchain. Each curator owns their curation graph and controls reputation they accumulate, similarly to how they control tokens today. It is controlled by the same key.

Subjective Reputation ‘Mining’

By endorsing content (or backing other curators) users convert their token holdings into reputation generated for content & other curators.

Reputation understood as a currency has different characteristics than money. Money is objective — if I give you a dollar and Mark Zuckerberg gives you a dollar the value you receive is the same.

Reputation is subjective. But if you get two Facebook likes — one from me and one from Mark then the difference in value received will be huge. Mark’s like will drive way more attention to you than mine. But that value will be realised only in the Facebook’s reputation economy. Your like received from Mark is worthless in Twitter’s reputation economy.

That’s because all of these platforms operate as silos controlling the entire closed stack.

In Userfeeds, all reputation transactions are objectively timestamped on the blockchain. Identities, token balances and their respective curation graphs are publicly accessible.

Each application can then access this graph and introduce their own subjective reputation ‘mining’ algorithm that will be used for ranking content in the context of this application.

If the ranking algorithm proves to be valuable for users, they’ll pay attention and drive traffic to that particular application. It will increase the value of the underlying, objective token and the subjective, app-level reputation currency. Reputation currency is a derivative currency based on the token, that only exists in the context of the app. We will explain the reputation currency model in more detail in our future posts.

Network Effect Hypothesis

Our hypothesis is that reputation tokens used purely for content ranking will increase in monetary value due to their ability to attract attention.

The feedback loop is as follows:

the token has no monetary value outside of reputation in a particular distributed interest group.
due to activity of token holders (endorsements, backings) in a given application, the application starts to attract attention (traffic).
new joiners can either get reputation from backers or just buy the token themselves. Some of them decide to buy it on the market.
demand for token increases, the price goes up.
early curators get rewarded, the higher price drives more attention.
a circular economy gets kickstarted which is based on exchanges of value between three scarcities: attention, reputation and money.

Why It’s Not a Decentralised Social Network.

It is not our goal to build a new, decentralised social network on the blockchain.

The vision of Userfeeds is to provide a token-based discovery layer that will enable curation of the existing, centralised social networks. Instead of building new Twitter, we want to allow users to curate the existing Twitter using the token. And Twitter is just the start because our reputation model applies to other content on the Web as well.

Of course, once the decentralised social networks emerge, you’ll be able to curate those as well. However, our perspective is that tokens are essentially new socio-economic networks with their own particular dynamics.

Today, the majority of interactions in these networks are money-based. With Userfeeds, we’re enabling a new set of reputation-based interactions (such as endorsements, backings and others).

How Curators Will Make Money?

The question how to reward curators often comes up in discussions about Userfeeds.

Our model doesn’t contain any protocol-level, direct reward schemes for curators.

The protocol itself is very simple — token holders timestamp transactions on the blockchain that off-chain applications interpret as reputation transfers. Reputation is calculated individually in each app.

This app-level reputation becomes valuable proportionally to the attention the applications will attract (represented as traffic, daily active users etc). Reputation in low usage apps will be worth nothing but it will grow in value along with user interest. It’s also possible that the same user will be high-reputation in some apps and low-reputation in other apps.

There are many possible scenarios where high-reputation curators can convert reputation into cash.

Let’s analyse one hypothetical scenario:

A curator with high ETH reputation accumulated by curation of tweets launches a new token called ‘Promoted’ along with a new application with a sleek user interface. She keeps 100% of the issuance to herself. Aside from her activity using the ETH token she starts endorsing content with her ‘Promoted’ token.

When approached by people who want her to endorse or sponsor their links she just sells them a fraction of her ‘Promoted’ token. This way they can curate the ‘Promoted’ timeline in her application. Or in other words — they get access to her ‘Promoted’ timeline. Just like FB placing ads on your Newsfeed but in a peer-2-peer way and with full transparency.

Our curator picks and chooses who she sells the token to in order to maintain the quality of the app. She understands that if ‘Promoted’ gets sold to everyone it may generate a quick profit but in the long run will decrease her reputation.

In this case, Userfeeds is used as an open, transparent advertising network where identities, ‘ad-buys’ and prices are permanently visible to everyone.

Imagine if you could look behind the banner on a website, check the identity of the advertiser and the history of all their transactions.

Userfeeds can be used as a reputation-only, money-free attention economy or a greedy, capitalistic advertising network. On the protocol level, both cases look exactly the same. Users voting with their attention and reputation transactions will give value to either one of the use cases.

Game Theoretical Considerations

This brings us to the important issue of incentive compatibility of our model.

According to Wikipedia incentive compatibility means:

“A mechanism is called incentive-compatible (IC) if every participant can achieve the best outcome to him/herself just by acting according to his/her true preferences.”

In our context this means that actors in the system should be rewarded for behaving truthfully, revealing their true preference without strategising and gaming the system.

We achieve that by introducing multiple levels of curation and various roles counterbalancing each other for the competition for scarce attention of users.

Roles and goals:

Curators aim to curate content and other curators in a way that maximises their reputation as calculated by applications.

Applications aim to optimise their ranking algorithms in ways that reward the best curators with the highest reputation.

Users aim to optimise their attention by choosing apps that give them the best Return-On-Attention (the best informational value for their time spent).

Interaction types

reputation transactions — on chain, consensual, objective — public
reputation scoring-— off-chain, app-specific, subjective (ranking algorithm) — public or private (app operator decides whether to make their algorithms public)
attention transactions-— off-chain, app specific, objective (clicks, impressions, etc), mostly private (visible only to the app operator)

Incentive compatibility

Each reputation transaction is recorded on the blockchain forever. At the same time, reputation scoring is evolving along with the applications’ increasing intelligence.

Curators can try to game the system for the short term increase in rankings and attention. However, in the long run, such behaviour will be noticed and penalised by the future applications with more evolved ranking algorithms.

Curators playing the long game have to optimise for filtering heuristics that are unknown today.

All we know is that the future apps will be exponentially more intelligent and that competition for attention will get fiercer (due to exponentially growing supply of information and constant supply of attention).

How should the curator behave today not to get filtered out tomorrow?

To answer this question, we need to guess what AI-enabled apps representing their users will consider attention-worthy in 10 years time.

We can only hypothesise about that today. Our assumption is that all predictable, average, repeatable patterns will get filtered out. In our view, the best strategy for curators is to reveal their best behaviour and avoid engaging in game playing and abuse.

This is in stark contrast to how the Internet is organised today, where the entire attention economy is optimised for short-term extraction via clickbait, sensation and scandal.

From Organising World’s Information To Allocating World’s Attention.

Google’s early mission statement was to ‘organise the world’s information and make it universally accessible and useful.’

In 2016 we can say that Google has succeeded in their mission. But along the way a new problem has emerged. Given that I have all the information in world available to me, what information I should pay attention to? How should I allocate my scarce attention knowing that there’s an opportunity cost of not paying attention to something else.

Today the decisions of world’s attention allocation are made by algorithms controlled by a few for-profit companies using closed algorithms to meet their quarterly revenue targets. While most people don’t see the link between the stock market and their social media feeds, in reality, this relationship is extremely strong. Economic incentives influence the proprietary algorithms used for ranking information. That information then shapes the worldview in our minds.

We believe that you can’t change this system without changing the underlying economic incentives it operates on.

The blockchain technology gives us a tool to experiment with new incentive models for algorithms that allocate our attention. With Userfeeds we build a conceptual framework and a technological platform to experiment with these algorithms.

Think Big. Start Small.

The cryptocurrency community understands how inadequate the existing content ranking algorithms can be when faced with economic incentives introduced by tokens.

Reddit brigading, trolling, spam or Twitter fights between prominent figures happen on a daily basis. Every single issue that could be decided in a day, is debated forever.

So for our first experiments, we plan to target the existing communities — starting with Ethereum. Ethereum already has many token communities so whatever we build immediately applies to all of them. And everything we learn with the Ethereum token, we can then quickly apply to communities operating on other blockchains starting with Bitcoin.

First Alpha Prototype — Elect Twitter Representatives With Ether.

For the first prototype, we’ll focus on Twitter.

Ethereum token holders will be able to elect their Twitter representatives and create a ‘Twitter feed backed by Ether’. A simple demo of how that might look like is available in the video (please note that the UX will definitely change before the alpha release on testnet — we just want to give you a general idea).

We’re adding some new features and will be ready for the testnet launch soon. If you are interested in joining the alpha test — please sign up for the alpha test.

If you have any feedback or comments — please lets know down in the comment section.

Credits:

We’d like to say thank you to people listed below for their contributions, inspiration and feedback. Each of them contributed in their own way and was substantial to the current shape of Userfeeds. While Kuba is pursuing other projects, Meher and Simon are actively exploring the blockchain ‘attention economy’ space from their angles — check out their Twitter and Medium for details.

Further Reading:

Liquid Democracy — the Wikipedia entry

Google PageRank- the original whitepaper

Michael Goldhaber’s article on the attention economy.

Albert Wenger — World After Capital-the book

John Hagel — Return on Attention and Infomediaries — the blog post

my appearance on the Epicenter Bitcoin podcast discussing Attention, Reputation and Money economies.

…and my writings on my blog and here on Medium