Signal Distortion Is by Design

The internet provides a stage for almost everyone. People can express themselves all the time. This is new. Historically, the ability to contribute to the public record belonged to a select few. You needed access to a press or a television channel, an institution, or money to transmit your information. A person could have noticed a dragon and had no place to put it. Now a person can notice a lizard and record it many, many times.

In aggregate, this should have become a continuous record of human experience. What is happening in San Francisco? Is this coffee shop open? What are people seeing in this war? How are people reacting to this policy? What actually happened to Real Madrid last season? These are very different questions, but they all depend on the same thing: a useful record of what people are seeing, experiencing, measuring, and reporting.

It would not be a perfect map of the world. No record is. But it could be good enough for all practical purposes. More than useful, it should be one of the driving artifacts of society. Governments, markets, institutions, communities, and individuals all depend on knowing what is happening.

We do not have a system like this. We do have systems that look very close to it. Their stated purpose is to connect people, enable open communication, allow people to express themselves, share opinions, form communities, and make meaningful connections, both professionally and privately.

But their operating objective is different. To understand it, you only need to look at how they operate, not what they say.

The Attention Economy

The objective is your attention. Attention is the coin of the digital economy, and much of Silicon Valley runs on this life force. Your attention is farmed. While you are paying attention, your behavior is observed. Every click, pause, scroll, return, abandonment, and reaction is information about you.

That information is used to build a predictive model of the user. The large digital corporations building these models also have extraordinary agency over the digital environment in which users act. They decide what appears, in what order, how often, and what comes next.

The prediction drives the experience. The user responds to that experience, and the response tells the system what worked and what did not. The system adjusts and tries again.

attention → information → model of the user → prediction → experience → behavior → more information

This is continuous experimentation on the user at enormous scale. The system is continuously learning what holds attention while continuously extracting more information about the person. The loop does not merely learn what the user already wants. The experience produced by the model changes what the user encounters, what the user becomes accustomed to, and eventually how the user behaves. That new behavior becomes new information, which improves the model, which produces the next experience.

The person becomes the model, and the model begins to shape the person.

The Selection Problem

There is another loop operating at the same time. The system also has to decide what to show. The selection parameter is attention.

Attention itself is not bad. Real things can capture attention. Important things can capture attention. Beauty can capture attention. The problem is that attention can be hacked.

"If it bleeds, it leads."

We do not react equally to everything. Fear, anger, jealousy, envy, lust, conflict, novelty, and outrage can provoke much stronger reactions than ordinary things. A forest can be beautiful. A pleasant conversation can matter. A useful observation can be important. None of them necessarily competes well with provocation.

Once attention becomes the selection criterion, the system has no reason to distinguish legitimate attention from attention elicited by provocation. If it captures attention, it satisfies the parameter.

selection → attention → measurement → improved selection

The selection improves continuously, but it improves toward what captures attention, including what provokes. Not toward accuracy. Not toward importance. Not toward a better account of reality.

Sometimes what performs is also true. Sometimes it is useful. Sometimes it is exactly what people should see. But that is not the selection condition.

What performs is not the same as reality.

The Signal Processing Error

This creates a signal-processing problem. People are both participants in the system and consumers of its output. We learn what gets attention and begin to express ourselves accordingly. At the same time, we look at what comes out of the system and treat it as information about the world.

Useful information can disappear simply because it does not travel. Provocative information can become abundant because it does. Then even ordinary questions become harder than they should be.

Why did a company fail? Bad leadership, a weak product, poor timing, a competitor's decision, a supply chain disruption, a change in consumer behavior, or a single event that tipped the balance. Reality may contain all of these at once. A simplified explanation may be easy to repeat. That is not a method for finding causes.

If the raw observations are gone and only the material that performed remains, we may no longer be able to reconstruct what actually happened.

AI Inherits the Record

We are now training artificial intelligence on that material.

Social media is only one part of the problem. Fiction, autobiography, journalism, books on economics, television, film, and other forms of recorded human expression are not neutral records either. They are written for purposes. They select, compress, dramatize, persuade, entertain, and provoke.

Entertainment makes the problem easy to see. A movie may take you through extreme happiness, extreme misery, fear, relief, excitement, and despair because strong emotional movement works. There is nothing wrong with entertainment. But entertainment is not the same thing as reality.

The problem begins when a record shaped by performance and provocation becomes the material we use to reconstruct the world.

Reasoning on top of a malformed signal does not repair the signal.

A Different Objective

The alternative is not to eliminate attention. Attention is useful. The alternative is to stop using provocation as the organizing principle.

A useful observation should not need to provoke everyone before it becomes visible. A contribution should keep its origin. A new observation should remain distinguishable from a repetition. Information should be placed where it matters, rather than promoted solely because it performs.

The record does not have to decide immediately what is true. It has to preserve enough of what was observed, by whom, when, and in relation to what, for truth to remain recoverable.

The internet made a continuous human record possible.

We should build for that.