Two people exchange one sentence. Two molecules collide. Two companies complete one transaction. All three are real interactions.
But we would not normally say that a new organization has formed.
That tells us something important: there are several steps between having causal contact and becoming a larger system that can operate through time.
Connection Theory uses a deliberately simplified path to separate those steps:
Interaction → Feedback → Connection → Organization
This does not mean every interaction eventually becomes organization. It asks a narrower question: what has to happen for a transient encounter to begin shaping the system's future?
Interaction: a real causal effect occurs
At the minimum, interaction means that the state of one thing changes what happens to another.
The effect may occur once and leave no durable trace.
A chance conversation can alter one decision without changing the later relationship between two people. A price fluctuation can change one transaction without creating a continuing supplier relation.
Interaction tells us only this:
an influence occurred.
It does not yet tell us whether that influence will matter again.
Feedback: consequences begin to return
If an action produces consequences and those consequences can affect later state or behavior, feedback has appeared.
A customer complaint influencing the next product revision, an experimental result changing the next hypothesis, and machine output altering later controller action are all forms of feedback.
But two limits should be stated immediately:
Feedback ≠ truth, and feedback ≠ adaptation.
A consequence returning to the system supplies new evidence or constraint. That evidence can be noisy or misinterpreted. A system can also respond with the same fixed rule every time, without changing the disposition on which future behavior depends.
Feedback is therefore important, but it is not yet organization—and it does not by itself prove that learning has occurred.
Connection: a causal path begins to enter the future
In this framework, a connection is not merely a line drawn on a network diagram. It is a causal relation with enough repetition, persistence, or predictability to become part of later dynamics.
A customer-support team repeatedly routing a class of problems to a product team is closer to a connection than a one-off message. Repeated synaptic influence is closer to a connection than a single accidental collision.
A useful intuition is:
An influence that happened once is an interaction; a pathway that repeatedly enters the future is closer to a connection.
But many connections can still form a disordered network. Their existence alone does not imply an integrated system.
Organization: relations begin to work together
We have stronger grounds for speaking of organization when stable relations are coordinated through time so that they jointly generate a persistent, distinguishable pattern of whole-system behavior.
Organization therefore asks more than “who is connected to whom?” It also asks:
- how information actually flows;
- which consequences can return;
- under what conditions relations are activated;
- how local processes coordinate;
- how one result becomes another process's input;
- how the whole maintains an empirically distinguishable relation to its environment.
A compact form is:
Once organization becomes meaningful, an inside–outside distinction can also become meaningful: which processes belong to the system's own operation, which come from environment, and which exchanges cross regulated interfaces.
A company example makes the difference concrete
Suppose customer support receives a complaint.
If one employee sees it and it disappears, that is an interaction.
If the consequence returns to the company and changes one later response, feedback has occurred.
If support repeatedly routes a class of problems to the product team, the path is becoming a connection.
If support, product, engineering, and management form a stable division of function—repeated problems are aggregated, checked, used in product decisions, and the resulting changes are tested again with customers—we now have more than one connection. We have part of an organized causal loop.
The example also shows why headcount does not create organization by itself. What matters is whether relations are coordinated into a system-level function.
Why the distinction matters
Without separating interaction, feedback, connection, and organization, it becomes easy to call every complex network a system, every feedback signal learning, and every increase in connection count an increase in capability.
Connection Theory asks a more discriminating question:
Which causal relations have actually become coordinated into a continuing whole that can receive consequences from reality and let those consequences alter its future?
So the most useful question is not merely whether connections exist. It is whether those connections are organized through time in a way that makes a persistent system possible.