Right Information is Liberation

Imagine trying to find that one specific book inside a gigantic library.

You could walk by each aisle looking at the sign boards of the categories hoping you will eventually land on the right shelf. But most people wouldn't choose that path—when information itself is abundant, the limiting factor becomes navigation instead of access.

To get to the right book you ask a librarian, who acts as a filter reducing your search costs and helping you achieve your goal (finding the book) faster.

The Evolution of Information Gatekeepers

When the information age grew, the librarians became the editors—major media platforms and other large media institutions. They classified these large volumes of information into categories and channels where people could access the right information.

Soon after, this created dependency on the media to serve the right information. Slowly the information in the media gets controlled by a tiny group of powerful individuals who can shape the views of thousands of lives—even without their knowledge.

"The control of information is always something the elite does." — Noam Chomsky

Then the internet revolution happened—it also led to the fall of such media's ability to control information flow and gatekeeping.

Even then, the problem of navigation persisted, arguably growing even harder. This time the librarians were platforms and algorithms. These platforms then tried to personalize a person's information in order to make them stay in the media platforms more.

Agents with Unaligned Incentives

Here we see another misalignment of incentives. People slowly start to fall in love with their recommendation systems as the systems learn more about these buckets of humans who prefer these over that. And slowly, using enough power, someone can influence social media to inject information and shift the masses toward their biases.

In a world with gargantuan levels of information, people seek guides to get their information.

In the US, Reuters Institute data shows 54% of 18–24-year-olds say social/video networks are their main source of news.

This growing dependency on recommendation algorithms can lead to echo chambers where everyone is fed different narratives about the world. This leads to polarity.

As a researcher at Facebook said: "Research after research showed that models that maximize engagement increase polarization."

Polarity is epistemic separation: different groups consuming different "facts," different framings, different emotional triggers—and then trying to coordinate a society as if they share the same baseline reality.

Polarity in opinions is often solved through debates or fact checks. Without these clashes of ideas and tests against one another, people retreat deeper into their own mental models of the world—reinforcing beliefs without meaningful challenge.

The Need for Better Personalization

So personalization itself isn't new. It has been present even from ancient times—one of the perfect examples is tutoring or home instruction. Scholars have highlighted how this direct master–apprentice relationship shapes the learner.

We know a world without personalization doesn't exist. It's when we offload the work of personalization to external agents who have misaligned incentives to our own goals that it leads to much broader negative impact.

Let's call this agent-based personalization with misaligned incentives "pseudo personalization." All of the current social media platforms and world's media currently do this pseudo personalization.

True Personalization

True personalization is where your data is used for your own benefit and prioritized higher than any other incentives.

This has been a huge challenge. Modern algorithmic feeds and search platforms are working hard to make the best algorithms at large scale so they can sell users products (misaligned incentives).

Personalization and recommendation often involves crunching through lots of data to pick up patterns about a user and usually classify them into buckets of users to whom content can be recommended. Doing it for each and every user runs into the problem of compute.

Half of why we optimize for the group has to do with the amount of compute it takes to do it at large scale and fast. So people found a way to automate this by training a big AI model on all user-item interactions so the AI can pick up patterns from this huge amount of data and use the same model to personalize for everyone.

For a world with limited compute, this would provide enough personalization. But with the trajectory of the world and hyper-scaling in compute, this kind of personalization looks like a lousy solution to an important problem.

The Right Information

What would we classify as right information? One which empowers the individual once they gain access to it. Getting the right information to the right individual is true personalization.

This problem is still wide open and hasn't been solved—because it is hard.

Just because it is hard doesn't mean it's impossible. As we see it, this might be one of the single biggest problems we will have to nail in order to empower each and every human.

Instead of algorithms controlling your life, the algorithms must learn you as a person truly and amplify each and every human's specialty. So everyone can truly be themselves.

---

The United Nations Global Principles for Information Integrity include the recommendation that tech and digital business models should not "prioritize engagement above human rights, privacy, and safety" and should give users more control.

Most of the 20th century—and even human history—this held true: the elite controlled information. But is it true for the modern age? In a world where any single human has the ability to find any information, how could the elite control information then?

The answer lies in navigation, not access. And navigation is what we're here to fix.