What if you could tweak the algorithm?

Where the money will be in decentralized social media

Welcome to the 5+ new people that have joined since last week! If you have not subscribed yet you can join this swarm 🐝 of smart and curious people for weekly insights.

Hello! Welcome back to Cloud Vertigo, a weekly newsletter on everything decentralized communication and more.

Today I am releasing What if you could tweak the algorithm? the second episode of the Unbundling Twitter series on decentralized social networks. If you have not read the first episode, you can find it here:

Unbundling Twitter

📚 This is what I have been reading this week:

• Why Decentralization Matter (2018). Venture capitalist Chris Dixon’s formidable and short post is a must read. I am heavily indebted to his argument

• Dan Wang’s 2022 Letter is out. An annual (very) long form by the geopolitical analyst who best understands China.

• Availability Cascades (1999) by Timur Kuran and Cass Sustain. Marc Anderseen substack post made me want to read this seminal paper

Since writing felt a bit hard and distractions were many, I occasionally called out a favor from Chat-GPT. How did it do? Judge for yourselves. Thanks as well to Dall-E for the illustrations.

What if you could tweak the algorithm?

Cupertino, 2033 — Apple released the latest version of News. It announces a new paid feature called Focused mode. It de-clutters your unified news + social media feed from all the ads, the promotional campaigns, the click-baiting titles, it shields you from viral reels and memes. It’s guaranteed to bring you the dullest shade of dull. But, just your kind of dull.

Close that cookie jar. Eat healthy.

Would you press that button if it was free? Personally, I would. What if you had to pay 1$ an hour? or 5$? At some price point most of us probably would not. After all, if highly addictive recommendations engines was what made social media big, why would anyone ever switch to a healthy and balanced news diet?

The journey of the first focused user

Imagine a world where you could customize your own Twitter algorithm. What if you could have more control over the content you see? To make what I mean more vivid, let me sketch a story of the first user that tries it and the lessons he learns along the way.

Our protagonist is a young professional, eager to make connections and expand his network on Twitter or LinkedIn. He spends hours scrolling through his feed, trying to find content that is relevant to his interests and industry. But he can't seem to shake the feeling that something is missing.

One day, our protagonist discovers that the algorithm is heavily weighted towards popular content and engagement, rather than relevance or quality. He realizes that this is why he is seeing the same types of posts over and over again (echo chamber), and missing out on valuable information and connections. His feed is not just personalized, but also optimized to give him cheap thrill that make him stay longer on the platform and make advertisers happy.

Determined to take control of his social network experience, our protagonist begins to experiment with his settings and preferences. At the end of a hidden panel he sees a dark red button, it says: “Do Not Press me”. The protagonist is very curious and presses the button anyway. He’s prompted to connect a Bitcoin wallet (or a credit card, but he does not own one) and he’s asked to set a hourly Focused budget. He does it.

The new Focused mode swiftly turns off notifications for certain types of content, suggests him to follow new accounts, and start engaging with posts that are relevant to his interests. Some of his allocated budget goes to reward quality publishers. Gradually, he begins to see a shift in the content that is presented to him, and a corresponding increase in the quality and relevance of his connections and interactions. After a while he’s bored and decides to go for a walk.

At the end of the day, he’s served a hefty invoice. The next day he disables the feature and goes to his mom (or his employer) to complain about what happened to him. When the mom hears the story she cannot believe it: “Can you really do that?!¨. She runs for her company credit card and adds an unlimited Focused budget to the her husband’s account!

The moral? Maybe we don't have to accept the algorithms that are given to us. We have the power to customize our experiences and shape our online interactions in meaningful ways. By taking control of our settings and preferences, with love and care, we can connect with the people and content that truly matter to us. So go ahead, put your health and productivity first, and tweak your LinkedIn/Twitter/Instagram algorithm - who knows what opportunities might be waiting for you on the other side?

Well, too bad we can’t. Not really, not yet.

The untold story of the algorithm

Don’t get me wrong. The algorithm is not the bad guy of the story. Today’s social networks are an easy target. They are so powerful that they seem to make everyone unhappy. A great deal of recent public debate has been focusing around how dangerous these platforms can be. Billions are spent moderating content. Many seek a more active approach from regulators. It would be a mistake.

Software-networks can powerfully evolve with entrepreneurship and market forces. There are unbound possibilities, software encodes human thought. Architectures can be redesigned. Bigger incumbents inevitably are slow and short-sighted.

In his “Why Decentralization Matters” Dixon’s point is very compelling. Encarta thrived where Wikipedia failed, even if at the beginning Encarta had a superior product, what mattered was the subsequent rate of change. Platform relationships with users and complements suffered a ¨bait and switch” fate.

But to every extent if #web2 social networks gained such traction and business, it's because they made some very happy. No, not just their shareholders. I am not even just referring to the advertisers who devoted more and more of their budgets to the platforms, where return on investment was higher and easier to measure. They also made users - or in last week issue’s “pub-sub” terminology - subscribers far better off.

Centralized platforms empowered connectivity at unprecedented scale. They also solved discoverability far richer and immersive experiences, for example compared to e-mails. TikTok’s strenght is precisely its recommendation algorithm (more novelty, more engagement). The new algorithm, the new Reel format made it rapidly gain market share with new generations.

Social networks are often criticized for their recommendation algorithms and how they personalize our news feeds. Many argue these algorithms are designed to keep us hooked and engaged on their platforms, while others believe that they are a necessary tool for discovering content we might not otherwise have found. Regardless of your stance, it's clear that recommendation algorithms have become the ubiquitous feature of the modern social media landscape. I am arguing that quality recommendations (as well as the cheap dopamine hit of a cookies sometimes!) are features that needs to be preserved in the upcoming #nostr based decentralized social networks. We just finally have the opportunity to give back some more control to users to “choose their own adventure”.

The key thesis of this chapter of the series is that a closer understanding of the underlying economics and the wider data availability of decentralized open protocols would empower the design of more efficient curation equilibria.

Last week, we saw how Unbundling the social media offer allows for better servicing of individual user needs. My hunch is that the biggest piece of the pie of the unbundled value propostion lies in content curation.

To sum up, the broader theme here is the potential for disruption that decentralized social networks represent. Just as blockchain technology is disrupting traditional financial institutions, #web3 has the potential to disrupt the centralized social media platforms that currently dominate the market. Its strength is structural.

Decentralized social network may not work

Today, social media platforms force many kinds of lock-ins: not only they lock-in your content and your audience, but also your reactions and activity (comments, likes, …). Some platforms, like this one, creators full ownership of content and audiences.

John Wolpert’s ¨Quit Twitter with Quitter” brings a good example of how to solve the content lock-in problem in a centralized social network without the need of decentralizing the whole platform. But content and audience is just the beginning…

Quit Twitter with Quitter

User activity seem of little value in itself, but is immensely valuable in aggregate, since it enables quality content and product recommendations (ads).

Critics could point out to all the features that are current missing in decentralized networks and the technical complexities of their infrastructure.

But good recommendations require network effects…

True, curation is a by-product of the subscribers' activity and their reward interactions. The more users on the platform, the more engaged they are, the more they browse and interact with publishers rewarding them, the easier it gets to accurately curate information flows. If there were zero interactions, there would be no data to make inferences from curation. If there is only one user, his interactions are worthless, when there is a , they become valuable. It's the task where network effects matter the most.

Lock-in is not inevitable to obtain network effects. With an open, extensible protocol - such as Nostr - you can have a greater degree interoperability. The best way to prevent information silos in decentralized communication is through interoperability. Interoperability refers to the ability of different systems or platforms to work together seamlessly. In the context of decentralized communication, interoperability allows individuals and groups to connect and communicate across different networks, making it easier to share and access information.

One of the main challenges of interoperability is that it requires agreement and coordination among different stakeholders. For example, different clients may support different NIPs improvement protocols, enforce various rules, and standards, which can make it difficult for them to communicate with each other. To address this challenge, projects are working on developing common standards and protocols that can be used across different client and relays platforms that are extensible in a way that maintains backward compatibility, with older clients. Maybe in the future, where it is strictly necessary, we might also see tokenised cooperation incentives?

But good recommendations require privacy trade-offs…

A degree of knowledge into my peers activity is very valuable for my content curation. It’s the fundamental Google’s intuition of the PageRank algorithm. Trust is transitive. If I trust Alice and Alice trusts Bob, I am likely to trust Bob. More trusted, or higher "rank" nodes carry more reputational value.

It’s also Facebook’s Social Graph. Trust is not just an absolute value, but also relative to who is trusting. In this sense we can better speak of peer similarity, or affinity. I am likely to agree with what my closest peers agree with, and their peers and so on. While this is simple principle has the potential to replicate echo chambers, it’s very important to incorporate some degree of similarity inference into a recommendation model.

Letting the user choose how much he wishes to be served accurate (specifically relevant) recommendations vs novel (generally popular) recommendation would be already a tremendous improvement. Exploring the use of decentralized reputation systems that allow users to rate and recommend content, while also providing visibility into how those ratings and recommendations are calculated, is a key future challenge for #nostr.

Overall, privacy in a protocol-based social networks is a subtle issue: on one hand, more personal information is publicly available, but most is in some pseudonymous form. What is your current take on privacy in Bitcoin?

Why decentralized network could win in the end

• Decentralized social networks have the potential to disrupt the centralized platforms that currently dominate the market. By leveraging open protocols and key pair identities (a signature #web3 technology), they can provide greater privacy, control, and diversity of experiences to users.

• The AI content recommendation process (the social algorithms) is the most powerful force that shapes our online #web2 experience. It has greatly evolved since 2010. It will keep evolving. Whether it is the obscure algorithms that determine what content we see on centralized platforms or the community-driven (and maybe paid-for!) of decentralized networks, the way we create, consume and interact with digital content is always ever relevant.

• The best benefit of all is that open protocols encourage builders to build. That’s what allows for faster compounding growth and scale of experience.
  
  Disagree? Send me a note! This week, I want to deep dive in the economics of content curation. Any reading recommendation is extremely welcome. Thanks for staying onboard in this learning journey.

  Onwards.