Interiority has a protective half and a developmental half, and the protective half alone is not enough. A person whose interior is shielded from surveillance and coercion but never actually developed has sovereignty in name only — the interior is private, but its conclusions are simply supplied by whoever reaches it first, whether that is a tradition, a platform, a demagogue, or a helpful AI. Real interior sovereignty is not just an undisturbed space; it is a formed capacity. Which means the framework has to say something about how that capacity is built — and the answer reorganizes what education is for.

From installing a worldview to building the capacity for one

The conventional aim of education, stated plainly, is often teach children the correct worldview. The framework’s aim is different:

Teach them how to form, test, revise, and own their own relationship to reality.

That is a shift from content to capacity, and it names a specific set of skills that a person needs in order to be the author of their own interior rather than a receiver of someone else’s:

  • distinguishing observation from interpretation;
  • understanding uncertainty, and being able to hold it;
  • recognizing persuasion, including the kind aimed at oneself;
  • learning how emotions shape judgment;
  • comparing competing explanations rather than accepting the first;
  • changing one’s mind without humiliation;
  • tolerating unresolved questions instead of forcing false closure;
  • understanding that truths are not all one kind — some are empirical, some relational, some moral, some existential — and that the tools that settle one kind do not settle the others.

That last one matters more than it looks: much interior confusion, and much manipulation, works by applying the standard of one domain to a question that belongs to another — demanding empirical proof of a moral claim, or dressing a moral demand as settled fact. A person who can tell the kinds of truth apart is far harder to capture. This is the capacity the model calls epistemic agency, and it is the developmental counterpart to the protection worked out in inference is not jurisdiction: one page keeps others from engineering your beliefs, this one builds your capacity to form them yourself. It belongs to the whole arc of a life, not to childhood alone, because the skill of revising one’s own relationship to reality is needed most in exactly the seasons of life when reality changes most.

The failure mode of an abundant-signal age

There is a specific reason this becomes urgent now rather than merely admirable. AI may make signal abundant while making interpretation increasingly dependent on intermediaries — the summary, the assistant, the model that tells you what the flood of information means. That combination is new. Prior scarcity forced people to interpret for themselves because no one could do it for them; abundance plus a capable interpreter removes the forcing function entirely. And if people stop developing the ability to form their own interior judgments, they arrive at a strange and dangerous condition:

Informationally rich, and ontologically dependent.

They have access to everything and the capacity to conclude nothing on their own — every judgment about what is real, what matters, and what to do routed through an intermediary they cannot check. That is ontological dependence, and it would be a severe failure of the whole framework, because it hollows out the interior the rest of the function works to protect. A protected interior that cannot think for itself is not a victory; it is a well-guarded vacancy waiting to be filled.

This is why formation is not a soft addendum to interiority but part of its core. The model’s own tools make the danger worse precisely by being helpful: the better the interpretation a system can hand you, the less you are pressed to build your own, and the more attractive dependence becomes. So an AI that respects interiority has to be built to strengthen the user’s epistemic agency rather than substitute for it — to show its reasoning, expose alternatives, mark uncertainty, and leave the concluding to the person — which is the same discipline the framework asks of AI as an information navigator rather than an epistemic sovereign. The goal of a system that forms interiors well is not a population that trusts it. It is a population that no longer needs to.