The Problem

We are not running out of information.
We are losing the ability to stay oriented within it.

The Crisis of Orientation

The horizon keeps expanding. We can see more, measure more, model more, and generate more possible explanations than any earlier generation. That should make navigation easier.

But a wider horizon is useful only if we can still tell where we are, what matters, what remains uncertain, and what should change when conditions change. When fog begins to accumulate, more horizon does not help by itself. Without the instruments needed to navigate, more information can arrive at the same time as less orientation.

"Every crisis narrative alive today — AI, environment, economy, information — is the same crisis wearing a different domain's clothing: a crisis of orientation."
— Arne Mayoh

There are several powerful narratives alive right now about a crisis that could end civilization as we know it. Superintelligence. Artificial intelligence more broadly. Autonomous AI agents. The information ecology. The environment. The economy. Even parts of natural science carry the same unease.

Each one names a different domain. The common problem may be more basic: whether human beings and institutions can remain sufficiently oriented while the systems around them become faster, larger, more interconnected, and harder to interpret.

Two expressions of it are sharp enough to name directly. Super-intelligence is the version already being discussed openly — a system capable of self-improvement outpacing any human or institutional ability to stay oriented to what it is actually doing. Super-Optimization is the older, quieter version, already running inside goal-directed systems too large to answer to a single person's judgment. A corporation can optimize successfully while the people inside it become progressively less able to see what the whole system is actually serving.

Neither problem is entirely new. What is new is the speed, scale, and density of the environment in which we are trying to remain oriented.

Layer 1 — Orientation

Orientation is not the same thing as having more information.

It is the ability to locate ourselves within a changing problem-space well enough to understand what is happening, what remains uncertain, what matters, and what direction of inquiry or action is still available.

A system can contain enormous amounts of correct information and still be badly oriented. It may know many facts while failing to recognise which facts are load-bearing, which assumptions have expired, which signals deserve attention, or whether its objective still fits the situation it is actually in.

This is the first distinction:

More information can expand the horizon. It does not, by itself, give us a compass.

Layer 2 — Continuity of Understanding

Orientation in the present depends partly on being able to recover the path that brought us here.

Continuity of understanding is the capacity for reasoning, context, assumptions, uncertainty, and learning to remain sufficiently connected across time and participants that later work can continue rather than restart.

When continuity breaks:

Continuity therefore preserves the path. But continuity alone is not enough. A perfectly preserved path can still lead in the wrong direction if conditions have changed and nobody recalibrates.

Layer 3 — What Matters Is Not the Same as What Is Supported

Modern systems are increasingly good at ranking information by confidence, probability, authority, or evidential support. Those are essential questions. They are not the only questions navigation requires.

A strongly supported fact can be almost irrelevant to the decision in front of us. An uncertain observation can be highly relevant if it bears directly on a load-bearing assumption, an emerging risk, or the difference between two competing explanations.

We therefore need to keep two questions separate:

Confusing those questions produces a subtle failure: evidence can accumulate while orientation gets worse. The system knows more, but becomes less able to discriminate what deserves attention.

Layer 4 — Calibration

Navigation requires more than remembering where we have been. It requires continuing correction.

Calibration is the process of adjusting confidence, interpretation, priorities, expectations, models, and action as evidence and conditions change.

A simple navigational cycle looks like this:

A conclusion can have been reasonable when it was reached and still become poorly calibrated later. Preserving it faithfully is not the same thing as preserving understanding.

Preserve what should persist. Recalibrate what reality has changed.

Layer 5 — How Orientation Breaks Down

The crisis is not singular. Several interacting failures can reduce our ability to remain oriented:

1. Interpretive Entropy

Information survives while the context, rationale, and relationships that made it meaningful become harder to recover.

2. Semantic Drift

Concepts shift across time and contexts until people appear to share language without sharing meaning.

3. Reconstruction Cost

Time and attention are spent rebuilding prior understanding instead of extending it.

4. Continuity Threshold

Fragmentation may become self-reinforcing when reconstruction consumes enough of the capacity needed to maintain continuity.

5. Relevance Failure

The system continues receiving information but loses the ability to distinguish what materially bears on the present question, risk, or objective.

6. Calibration Failure

Inherited conclusions, priorities, or models remain stable while the conditions that originally warranted them have changed.

Layer 6 — The Pressures Are Increasing

These failures are amplified by system-level pressures:

AI intensifies all five. It increases the speed and quantity of reasoning-like output, but greater capability does not automatically preserve continuity, improve relevance discrimination, or keep human judgment calibrated to what the system is actually doing.

The danger is not simply that machines become more intelligent. It is that the environment becomes harder to navigate faster than our instruments for orientation improve.

Layer 7 — Governance Requires Orientation

Current approaches rightly focus on:

But governance operates downstream of orientation. Rules can specify what must be done, disclosed, or prohibited. They cannot by themselves guarantee that the people applying them still understand what is happening, which evidence matters, why a decision was made, or when the underlying assumptions have changed.

Without continuity, governance loses its reasoning lineage. Without relevance discrimination, it can attend to the wrong signals. Without calibration, yesterday's correct rule can become today's poorly fitted response.

No institution can govern coherently if it cannot remain oriented to the system it is trying to govern.

Layer 8 — From Continuity to Navigational Coherence

The challenge is therefore larger than preserving records or reasoning alone.

Navigational Coherence depends on several connected capacities:

These are not guarantees of correct judgment. They are conditions under which correction, learning, and responsible action remain possible.

This is the larger problem Audited State investigates: not merely how to preserve understanding, but how to preserve the human and institutional capacity to navigate coherently as the horizon expands and conditions change.

Society Is Already Trying to Address This

The EU AI Act is a serious attempt to govern important parts of this problem through transparency, oversight, risk management, accountability, and human supervision.

That matters. But regulation can mandate processes, disclosures, and responsibilities. It cannot by itself ensure that the reasoning behind a decision remains recoverable, that attention remains focused on the right signals, or that inherited judgments remain calibrated when conditions change.

A system can satisfy disclosure requirements and still lose the thing that made a decision intelligible in the first place — the ability to reconstruct why it was made, what uncertainty remained, and whether those reasons still apply.

Disclosure has expanded — logging, reporting, published detail.
Legibility has not kept pace with it.

A disclosed risk that cannot be traced back to a stable object, or pointed at a party positioned to act, remains invisible in practice even when it is visible on paper.

→ Legibility
→ Read the paper: Legible Risk: Why Disclosure Isn't Enough

The same gap shows up inside institutions, not just between them and their regulators. When AI participates in analysis, drafting, or evaluation without a visible trace of intent and justification, an institution can keep operating while losing durable access to the reasoning behind its own conclusions.
→ Shadow AI

This is not a criticism of regulation. It is the gap regulation cannot close on its own because part of the problem is structural and navigational, not merely procedural.

→ How the EU AI Act maps onto this — structural alignment, not legal advice

The Other Side of the Loss

Everything above describes what is being lost. The same quantity, seen from the other side, is what stays unattained: the value that work could have built, had understanding been carried forward instead of rebuilt. It is easy to miss for the same reason the loss is: nobody counts it.

→ The Unattained Gains — what recovery would give back

Transition — From Problem to System

The problem is not lack of intelligence.

It is that intelligence, information, and optimization are scaling faster than the structures required to remain oriented to them.

To address this, we must move from:

Next: Architecture

We are scaling intelligence without yet scaling the instruments required to remain oriented to it.

Related Concepts

The challenge is not only to preserve what we know,
but to remain able to find our way as what we know changes.