I’ve been exploring the idea that agency may live less in the fantasy of independence than in the ability to choose the permeability of a boundary.
That sentence is abstract, but the idea is simple. Nothing remains itself by shutting out the world completely. A cell stays alive by allowing some things through its membrane and refusing others. A person becomes recognizable through an ongoing exchange with other people, places, and ideas. At my day job I recognize the ways in which a publisher gains influence when its work circulates, but loses agency if someone else gets to decide every term of that circulation.
A boundary, then, is not merely a wall. It is a decision-making surface.
That definition changes the question we should ask about openness. The question is not whether a system is open or closed. Almost nothing important is fully either one. The better questions are: What can cross the boundary? Under what conditions? Who gets to decide? Can those decisions change when the environment changes?
A recent work trip made me notice the personal version of this problem. Travel replaces familiar cues with ambiguous ones. Rooms, routines, and group dynamics all require more interpretation than usual. Even when I expect to enjoy a trip, that uncertainty can create some anticipatory static.
The revealing part is not that discomfort appears. That is ordinary. The revealing part is that the mind does not evaluate every new signal equally. I found that because of my personal psychology and trauma-dripping history, social friction can pass through my ego boundary immediately, while “welcome” and “belonging” are forced to wait outside.
That is how a protective boundary can become a distorted one. It still regulates exchange, but it admits the wrong evidence. The result is not safety. It is a system that has become less able to perceive the environment it is actually in.
A protective filter can become a source of error
Imagine arriving somewhere unfamiliar. A conversation takes a minute to find its rhythm. Someone is slow to reply. A group shares context you do not yet have. None of these moments is especially meaningful on its own, but uncertainty creates space for interpretation.
A defensive filter tends to resolve that uncertainty in one direction.
Small moments of friction are admitted as evidence that something is wrong. Kindness receives more scrutiny. An invitation might be dismissed as politeness. A warm conversation can be treated as an exception. The system is not closed to outside information. It is selectively permeable: signals that confirm caution move quickly, while signals of welcome face a higher burden of proof.
Technical systems fail in similar ways. A security control designed around an old threat model may block legitimate traffic. A classifier trained on skewed data may reproduce the skew while appearing objective. A process created to prevent one costly mistake may quietly create a hundred smaller costs after the original conditions have changed.
The common failure is not protection itself. Protection is necessary. The failure is allowing a protective mechanism to become unaccountable to new evidence.
Human beings are not software, and our inner lives cannot be reduced to engineering diagrams. But the comparison helps clarify the problem. The goal is not to become indiscriminately open. It is not to declare every room safe or every invitation sincere. That would simply replace one bad filter with another.
The goal is to notice when a boundary has stopped distinguishing between danger and unfamiliarity.
That distinction matters because unfamiliarity is unavoidable. Every new relationship, workplace, city, or idea begins outside the current model. If the model treats whatever it cannot yet classify as a threat, it protects continuity by preventing learning.
A boundary can preserve a system. It can also prevent the system from discovering that the world has changed.
Agency means participating in how the filter works
The usual response to this problem is to reach for control.
Think better thoughts. Act confident. Ignore the discomfort. Override the reaction and walk into the room as if nothing is happening.
This can work as a performance, but it does not necessarily create agency. It divides the person into the part issuing commands and the part failing to obey them. The boundary remains unchanged; it has simply acquired an internal adversary.
A more useful question is: what evidence am I allowing myself to receive?
That question does not require pretending discomfort is meaningless. Discomfort can enter without becoming a verdict on the entire environment. A prediction can be noticed without being mistaken for an observation. An unfamiliar place can remain unresolved for a while instead of being classified immediately as either safe or threatening.
Most importantly, welcome can count as evidence.
This sounds obvious, but defensive filters often require welcome to be overwhelming before they recognize it. A saved seat, a thoughtful question, a conversation that becomes easier, or a room that gradually feels more familiar may each seem too small to overturn the model. Yet those ordinary signals are how belonging is usually offered. It rarely arrives as a grand declaration.
Belonging is relational. No individual can manufacture it alone, and not every environment will provide it. But we can become more available to the moments when it is actually being offered. We can stop asking welcome to defeat every defense before we agree to perceive it.
That is a modest form of agency. It is not mastery over every reaction. It is participation in how reactions become conclusions.
On my work trip, nothing cinematic happened. I arrived. The city became less strange. Conversations found their rhythm. Anticipation gave way to contact, and the experience became enjoyable in the ordinary way many experiences do.
The ordinariness was the lesson. The environment did not need to prove itself all at once. The filter needed to update one signal at a time.
Healthy boundaries make trustworthy exchange possible
This personal example points back to the larger systems question.
In my work, I think often about the boundaries between publishers and AI systems. A publisher’s reporting, archives, judgment, and language can enter an AI product and become part of what that product knows and how it creates value. The exchange may return attention, attribution, payment, distortion, or nothing at all.
The important question is not whether publisher content should circulate. Publishing exists because ideas are meant to move. The question is whether publishers can participate in governing the terms under which their work crosses into another system.
Permission, provenance, pricing, revocation, and limits on use are all ways of making that boundary governable. They allow a publisher to distinguish an invitation from an intrusion. They also make a meaningful yes possible, because consent requires the ability to say no or to change the terms later.
Organizations face the same tension. A team with no boundaries becomes chaotic. Priorities change with every request, responsibility becomes unclear, and attention is consumed by whoever asks most loudly. But a team whose boundaries reject every outside signal becomes brittle. It protects its process at the expense of its purpose.
The choice is not between total openness and total closure. In people, organizations, and technical systems, both extremes destroy agency in different ways. Exposure lets the environment make every decision. Isolation prevents the system from receiving the information it needs to adapt.
Healthy boundaries do something harder. They remain coherent while allowing trustworthy exchange.
I used to imagine that skill as mastery: a strong hand deciding exactly what gets through. Stewardship now feels closer to the truth. Stewardship means respecting why a boundary exists, noticing when its model no longer fits the present, and updating it without abandoning its protective purpose.
The border does not have to disappear.
It has to become capable of recognizing an open door.