The Future of Intelligence Depends on Who Gets to Say What Experience Means

What if the most valuable knowledge in the future is not the knowledge an organization can collect, but the knowledge a person chooses to reveal?

A former hostage standing at a university campus with a sign that says, “I survived 505 days of Hamas captivity. Ask me anything,” and a technology executive describing a firm whose assets are transformed into a “compounding cognition loop” appear to belong to different moral and intellectual worlds. One concerns testimony after unimaginable suffering. The other concerns artificial intelligence, workflows, and the future balance sheet of companies.

Yet they meet at a consequential question: What happens when lived experience becomes an interface for other people’s learning?

The answer determines whether the knowledge economy becomes a system of genuine collective intelligence or merely a more sophisticated machine for extracting value from human beings.

The New Asset Is Not Information. It Is Interpreted Experience

The old firm accumulated relatively stable things: employees, software, processes, customer relationships, data, intellectual property, and brand. The emerging firm aims to own something more dynamic: a system that observes work, learns from decisions, improves its outputs, and feeds those improvements back into future work.

In that system, every workflow can become a training surface. A customer complaint teaches the company how to respond. A failed forecast teaches it which assumptions were weak. A skilled employee’s judgment can be translated into patterns that software may reproduce. Each interaction potentially makes the next interaction better.

This sounds like a straightforward productivity revolution. But the phrase “training surface” quietly changes the status of everything that happens inside an organization. A conversation is no longer only a conversation. A decision is no longer only a decision. A mistake is no longer only a mistake. Each becomes raw material for the organization’s memory.

That creates a powerful advantage, but also a moral hazard. The more efficiently a system converts experience into institutional intelligence, the easier it becomes to forget that experience belongs first to someone who lived it.

The former hostage’s campus stand offers a striking countermodel. He does not simply allow his experience to be absorbed into the public record. He creates a deliberate setting in which people may ask questions, and he decides to answer. The stand is not a passive data source. It is a human being establishing the terms under which his experience becomes knowledge.

This distinction matters. Knowledge is not merely what can be extracted from experience. It is what a person has chosen to make available, in a form that preserves meaning and agency.

The future belongs not to the organizations that collect the most experience, but to those that learn without treating people as raw material.

The Difference Between Exposure and Agency

Consider two ways of learning from a difficult event.

In the first, an institution records everything. It captures a person’s words, emotional reactions, timing, choices, and mistakes. An algorithm detects patterns and turns them into a procedure. The institution becomes more capable, but the person may have no say in how the story is interpreted, who receives it, or what conclusions are drawn.

In the second, the person participates in the construction of meaning. They decide what to disclose, explain what outsiders might misunderstand, correct simplistic interpretations, and refuse questions that cross a boundary. The audience still learns, perhaps deeply, but learning occurs through a relationship rather than an extraction pipeline.

The difference is agency.

Agency is often treated as a soft ethical consideration, something added after efficiency has been optimized. In reality, it is a condition of high quality learning. Without agency, experience is easily distorted. Observers may confuse survival with preference, adaptation with endorsement, or a forced decision with a rational strategy. Context disappears precisely when the system becomes most confident.

A machine can identify that a particular action preceded survival. It cannot automatically know whether that action was wise, coerced, lucky, or the least terrible option available. It can observe what happened. It cannot, by observation alone, determine what the event meant to the person who endured it.

This is true in businesses as well as in extreme human circumstances. Suppose a sales representative consistently closes difficult deals. A company may record the representative’s emails, calls, and timing, then build an automated sales assistant. But perhaps the representative succeeds because customers trust her judgment. Perhaps she notices a hesitation that no transcript captures. Perhaps she knows when not to push. If the organization extracts only visible behavior, it may copy the surface while losing the source of effectiveness.

The same problem appears in emergency medicine, aviation, education, and military operations. A team that wants to learn from a crisis cannot simply gather logs. It must ask people what they saw, what they feared, what alternatives they considered, and which constraints shaped their choices. The most important variable is often absent from the record: the person’s model of the situation at the time.

A compounding cognition loop becomes genuinely intelligent only when it can distinguish action from intention, correlation from explanation, and outcome from judgment.

Why Testimony Is a Form of Infrastructure

Public testimony is often described as storytelling, but that description understates its social function. Testimony is a form of infrastructure. It gives a community a way to carry knowledge that no individual could independently acquire.

A person who has endured an extraordinary situation can provide others with a map of reality they could not have drawn themselves. What does fear do to attention? How does uncertainty alter time? Which forms of help are actually useful? What kinds of language comfort, and which kinds wound? These are not merely facts. They are orientation tools.

But infrastructure has design choices. A bridge can connect people, or it can be built in a way that excludes them. A database can preserve memory, or it can make sensitive information available to those who should never possess it. A public question stand can invite understanding, or it can turn suffering into spectacle.

The ethical quality of the exchange depends on whether the person remains an author of the knowledge. The stand’s most important feature is not only that questions are allowed. It is that the survivor is visibly choosing to be there, choosing the format, and choosing the terms of engagement.

That offers a useful principle for organizations building learning systems: every training surface should have a corresponding consent surface.

If a workflow generates data, who knows it is being used for training? If an employee’s judgment is converted into an automated recommendation, can that employee inspect the result? If a customer interaction becomes a model improvement, what rights does the customer retain? If a failure is transformed into a lesson, who decides whether the lesson is accurate?

Consent does not mean asking for permission once in a long legal document. It means making the path from experience to institutional memory visible and contestable. People should understand what is being learned, how it will be used, and how they can challenge a misleading interpretation.

This is not bureaucracy for its own sake. It is a way to protect the quality of the loop.

The Three Layers of a Compounding Learning System

A useful way to evaluate any cognition loop is to separate three layers that organizations often collapse into one.

1. Capture

What happened? Which actions, words, outcomes, and environmental conditions were recorded?

Capture creates the raw material of learning. It is also where privacy risks begin. An organization that captures everything may believe it is becoming more intelligent, when it is actually accumulating unstructured exposure.

2. Interpretation

What did the event mean? Which constraints shaped the decision? What did the participant know at the time? What would they change, and what would they repeat?

Interpretation is where human testimony becomes indispensable. Data can reveal that a decision occurred. It rarely explains the lived landscape in which that decision made sense.

3. Authorization

Who is allowed to use the lesson, for which purpose, and under what conditions?

Authorization is the layer most likely to disappear in a race toward automation. Yet without it, learning becomes appropriation. An organization may be able to improve from an experience without having the moral right to use it in every conceivable way.

These layers produce a simple diagnostic:

A system is not truly intelligent when it can remember everything. It is intelligent when it knows what to remember, how to interpret it, and what it is permitted to do with the result.

Imagine a hospital reviewing an emergency procedure. Capture gives it the timeline. Interpretation includes the nurses’ perception that a warning sign was visible but culturally difficult to raise. Authorization determines whether the lesson is used only for internal training, shared with other hospitals, or turned into a commercial product.

Most organizations are investing heavily in capture and increasingly in interpretation through machine learning. The neglected layer is authorization. That neglect will become more costly as systems grow capable of learning from every interaction.

The Paradox of Making Experience Scalable

Scaling knowledge is attractive because human experience is scarce. One person can endure an event, solve a problem, or develop a subtle skill. A system promises to distribute the resulting insight to thousands of people.

But scale creates a paradox. The more widely an experience is distributed, the easier it is to detach the lesson from the person who gave it meaning. A story becomes a slogan. A judgment becomes a rule. A survival tactic becomes a recommended procedure. What was once context sensitive becomes universalized.

The solution is not to keep experience private. It is to scale context along with content.

When a company converts expert behavior into software, the system should preserve uncertainty, exceptions, and conditions of use. Instead of saying, “Do this,” it might say, “This approach worked under these conditions, with these signals present, and the expert rejected it when these other signals appeared.” A mature learning system does not only produce answers. It exposes the boundaries around answers.

The same principle applies to public testimony. A survivor’s account should not be reduced to a motivational lesson about resilience. Resilience without context can become a demand placed on other victims: recover visibly, educate others, make your pain useful. The person’s experience must remain more than the benefit an audience can derive from it.

This is why the voluntary act of asking and answering is so powerful. It preserves the distinction between being useful and being used. A person may choose to make their experience useful without surrendering ownership of it.

Organizations need the same distinction. Employees should be able to contribute to collective intelligence without becoming invisible components of an extraction machine. Their expertise can improve the system while their authorship, credit, boundaries, and right to dissent remain intact.

Building Loops That Deserve to Compound

The practical challenge is to design learning systems that improve with use while preserving human authority. Several principles follow.

First, make learning visible. Tell people when their actions, conversations, or decisions are being used to train a system. Hidden learning may be efficient, but it destroys trust and prevents correction.

Second, preserve provenance. Every important institutional lesson should carry information about where it came from, under what conditions it was produced, and who can challenge it. A recommendation without provenance is an orphaned conclusion.

Third, reward interpretation, not just output. If teams are praised only for speed and measurable outcomes, they will feed the system clean data while suppressing ambiguity. Build rituals where people explain why a decision made sense at the time, including what they did not know.

Fourth, create refusal rights. People must be able to say that an experience is too personal, too sensitive, or too context dependent to be converted into a reusable asset. A loop that cannot accept refusal will eventually train itself on fear.

Fifth, measure the quality of the loop by the quality of future judgment, not merely by the quantity of captured information. The aim is not a larger archive. The aim is wiser action.

Key Takeaways

• Treat experience as authored knowledge, not free raw material. Ask who is contributing it, who interprets it, and who controls its reuse.
• Pair every training surface with a consent surface. Make data collection, model training, and institutional learning visible to the people involved.
• Separate capture, interpretation, and authorization. Recording an event does not explain it, and understanding it does not grant unlimited permission to use it.
• Scale context with content. Preserve uncertainty, constraints, exceptions, and the conditions under which a lesson worked.
• Protect the difference between being useful and being used. Voluntary contribution, credit, boundaries, and refusal rights are not obstacles to intelligence. They are its foundation.

The deepest lesson is not that human experience can become an asset. It already has. The question is whether the person who lived it remains present in the asset’s creation.

A compounding cognition loop can make an organization faster, more adaptive, and more capable. But if it learns by quietly converting every human moment into property, it will compound something else as well: distrust, distortion, and moral debt.

The better model is not a company that owns everything its people experience. It is a community that becomes wiser because people are willing to teach it, and because they remain free to decide what teaching means.

The most advanced learning system, then, may not be the one that asks every question. It may be the one that has earned the right to receive an answer.