Causal Boundary

In plain language: What your org can Reach.

Definition

An Agent’s Causal Boundary is the collection of all the ways the Agent can affect its surroundings. It is the output side of the Agent — the perimeter through which the Agent’s effects flow outward into the world. The Boundary is defined by what the Agent can change: a person’s physical actions, speech, writing, and decisions enacted; a single biological cell’s cytokine release, migration, and signaling; an LLM’s generated text and tool calls.

What the Causal Boundary is not:

• Not what arrives at the Agent’s senses or channels — that is the Computational Boundary.
• Not what the Agent models or understands — that is the Cognitive Boundary. The Causal Boundary and the Cognitive Boundary are independent: an Agent can affect things it does not understand (the Franz Ferdinand Effect), and can understand things it cannot affect. A decision — the act of choosing — is Cognitive; acting on that decision is Causal.
• Not limited to what the Agent intends. The Causal Boundary is the Agent’s full reach — every effect the Agent can produce, whether the Agent chose it, anticipated it, or has any ability to model it. An involuntary muscle spasm that presses a button is inside the Causal Boundary. An executive’s departmental reorganization that accidentally breaks a productive mentoring relationship he never knew about is inside his Causal Boundary — he can cause it, even if he didn’t intend it.
• Not tied to a moment or context. The Causal Boundary is treated as time-and-context-free — the reach surface the Agent has by virtue of being that Agent, considered independently of what that reach is doing right now. Capacity (the maximum effect the Agent could produce) and Realized (what the Agent actually affects in a given context) are perspectives on the same Boundary, not alternatives to it.

Relations

One of the three Boundaries every Agent has, jointly definitional with the Computational and Cognitive Boundaries. Information enters through the Computational Boundary, becomes understanding inside the Cognitive Boundary, and flows back out through the Causal Boundary as effect on the world. Causal-side Tools — a gun, a lever, a broadcast antenna — extend the Causal Boundary by amplifying or redirecting the Agent’s effects. The Causal Boundary splits into two sub-forms: Intrinsic Causal Boundary (what the Agent can do by its own power — physical strength, voice, innate capabilities) and Positional Causal Boundary (what the Agent can affect because of where it sits in a Composition’s Topology — a VP can restructure headcount because the org-chart position carries that authority — remove the person and the authority stays behind). Actual Output is the instantaneous effect flowing through the Causal Boundary at a particular position and time — the Causal Boundary is the reach surface, Actual Output is what is flowing through it right now. Capacity and Realized are the two perspectives: Capacity is the maximum effect the Agent could produce; Realized is what it actually affects in a given context. The Causal-to-Computational Interface names how Agents couple: one Agent’s Causal output lands inside another Agent’s Computational input. This is where the Causal Boundary meets the outside world — and specifically, the next Agent’s Computational Boundary. Communication is the alignment of these interfaces in both directions. The visualization vocabulary applies — pinched ellipse (a CEO at work actively exercises both Intrinsic and Positional reach; the same person on vacation retains both but engages primarily Intrinsic — the Positional reach is un-exercised, not absent); time-slice continuity (the Agent persists; its Causal Boundary reshapes as its position, authority, and tools change).

Example — CEO

A VP of operations reorganizes his department — new reporting lines, team moves, budget reallocations. His power to make these changes is positional: it comes from where he sits in the org chart, not from his physical capability. One person’s decision restructures forty people’s work because the Topology confers that reach.

The reorganization produces effects he intended — a new team aligned to a strategic initiative — and effects he did not. Two engineers had an informal mentoring relationship that was holding a critical knowledge transfer together. When one is moved to a different team, the relationship breaks and the knowledge transfer fails. The VP’s Causal Boundary includes the move (he made it). His Cognitive Boundary does not include the informal relationship (he had no idea it existed). This is the Franz Ferdinand Effect at the organizational scale: the Agent’s Causal reach exceeds its Cognitive modeling, and the gap produces consequences the VP cannot predict.

What his Causal Boundary does NOT include: the board’s strategic direction (the VP has no mechanism to affect it), a competitor’s pricing decisions (outside his company entirely), and any effect that requires authority or capability he does not have. The Boundary is defined by what the VP can actually change, not by what he cares about or wishes he could change.

The Intrinsic / Positional split is visible here. The VP’s ability to speak, write, and persuade is intrinsic — it travels with him wherever he goes. His budget authority and his power to reassign headcount are positional — they belong to the position in the Topology, not to the person. If the VP leaves the company, the Agent loses that positional Causal Boundary (because he is no longer in that company, so he cannot cause the same results), but the company’s Topology retains it: the next VP inherits the same positional reach, or the CEO stretches to cover the node temporarily, reshaping their own Boundary to fill the vacated position.

Frame applies to the Causal Boundary: the VP’s self-model of his reach can be wrong in either direction. He may think he can influence the board’s strategy when he has no mechanism to do so (Frame larger than Boundary). He may not realize his budget decisions shape the company’s talent trajectory for the next two years (Frame smaller than Boundary). The mismatch is structural.

Example — Research

A single biological cell’s Causal Boundary is the collection of effects it can produce in its environment. It releases cytokines that recruit other immune cells to a site of infection. It migrates toward a chemical gradient. It secretes enzymes that remodel the surrounding tissue. Each of these is the cell affecting its surroundings: Causal output.

What the cell cannot affect: a cell in a distant tissue with no pathway to receive its signals. The cytokine diffuses only so far, and whether a distant cell has a matching receptor determines whether the signal lands. But that is a coupling question — the Causal-to-Computational interface between two cells — not a question about this cell’s Causal Boundary. The cell’s Causal Boundary includes its full secretory and locomotive apparatus. Whether the output connects to another Agent is about the interface, not the reach.

The Intrinsic / Positional split applies at the cellular scale. The cell’s ability to release a specific cytokine is intrinsic — the molecular machinery is part of the cell. Its ability to influence a specific neighbor is positional — it depends on where the cell sits in the tissue, what is in diffusion range, and what receptors neighboring cells express. A cell that migrates from one tissue location to another changes its Positional Causal Boundary without changing its Intrinsic — same machinery, different reach.

The Franz Ferdinand Effect at the cellular scale: an immune cell follows a chemical gradient toward a site of infection — its receptor programs respond to the gradient and produce directed migration. As it moves through the tissue, it secretes enzymes that break down the surrounding matrix to clear a path. Those same enzymes damage healthy tissue along the route. The cell’s signaling programs determined the direction of movement. The collateral tissue damage is an unmodeled consequence — inside the Causal Boundary (the cell produced the enzymes and moved through the space), outside the Cognitive Boundary (no machinery to model the downstream tissue effects of its own transit).