Navigability turns on one principle — support orientation without owning destination — and this page is about everything that would violate it. The threat is not that people lack direction but that something powerful supplies it for them and calls that help.

The greatest navigator ever built

AI may eventually know a person’s capabilities, history, resources, health, preferences, relationships, available projects, and the economic conditions around them — and produce astonishing recommendations: given everything I know, this path has the highest probability of fulfilling your stated priorities. Useful, and extremely dangerous, because recommendation can become destiny by repetition. If everyone follows highly optimized recommendations, a society converges on what existing data already understands, and the horizon shrinks while appearing enormous. So the model requires that AI navigation preserve uncertainty, alternatives, unexplored options, its own assumptions, model disagreement, the explicit recognition of unknowns, and the participant’s ability to reject the recommendation. The system should say here are several plausible bearings and what each appears to imply — never this is your correct path. This is the emergence principle applied to direction:

Prediction must not consume the ecology’s ability to surprise itself.

It is the same line the model draws for AI everywhere — an information navigator, not an epistemic sovereign — carried into the question of where a life or a society is heading.

The same discipline binds government. A society can legitimately establish broad collective trajectories through democratic and constitutional processes — build renewable energy, invest in public health, expand transport, pursue research. But the state should be extremely cautious about assigning individual destiny in service of those goals. Civilization can choose directions; individuals retain substantial jurisdiction over how they participate in them, and that separation protects both coordination and freedom. It also bounds the individual in the other direction: a chosen bearing gains claims on the field only where competing standings and consequences permit —

You own your bearing. You do not automatically own the road, resources, or other participants required to travel it.

Someone may privately want to become the world’s greatest tyrant; the ecology has no obligation to supply the resources or jurisdiction that would require. Direction is the participant’s; the means to impose it on others is not.

Captured navigation, and the frozen compass

Two failures deserve to be named precisely, because both can operate on accurate information. The first is captured navigation — the propaganda problem, distinct from corrupted signal. Signal corruption says you don’t know what is happening; directional capture says you may know exactly what is happening, but someone has monopolized the interpretation of where you should go because of it — therefore support us, therefore your nation has only one possible future, therefore this group is the enemy, therefore history inevitably moves our way. So the model needs protection not only against false information but against monopoly over trajectory interpretation, such that a participant can always say I accept the facts and draw a different directional conclusion.

The second is the frozen compass. Any worldview — a religion, an economic system, a political philosophy, even this framework — can originally provide excellent bearing. The danger appears when the compass is treated as correct regardless of changed terrain: navigation stops, and doctrine replaces orientation. A tradition becomes doctrine exactly when it stops navigating and merely preserves. The framework has to hold itself to this too, which means the governing question is always given where we actually are now, does this direction still hold? — never what direction did our founders establish? And because no one can know the future, navigation, like signal, needs plurality: a society should keep a portfolio of future hypotheses — competing research programs, political philosophies, cultural visions, forecasting models, communities trying different arrangements — because a society with only one authorized future is extraordinarily brittle. Some hypotheses will fail; that is how the ecology learns.

Purpose is not bearing; bearing is not a pathway

Finally, two distinctions keep the function from being confused with its neighbors. Purpose is not bearing. What matters? is not the same question as where should we go? Suppose preserving human dignity matters — that still does not settle whether the best immediate direction is to expand AI access, slow AI development, invest in education, or strengthen local communities. Values constrain navigation; they do not determine the route. Bearing is the practical directional relationship among values, conditions, signal, commitments, and horizon — which is exactly why it needs its own function rather than dissolving into “what matters.”

And bearing is not a pathway. You can know precisely where you want to go and have no viable road: I want to become a physician is bearing; medical education, licensing, clinical training, and institutional access are pathway. The model’s navigation helps establish and continually correct direction; it does not guarantee the route exists — bearing without traversability is aspiration. Building the roads is the work of the next function, traversability, and keeping the two distinct is what stops the model from mistaking a clear destination for a reachable one.