The Cognitive Boundaries Framework uses a small vocabulary, but each item and how it interacts with the others is important. Every term below is defined and has examples for both CEO and research audiences.
The Three Boundaries
Every Agent (an entity, whether human, AI, a cell, or a collection of these) has three Boundaries. Together they define the category.
- Computational Boundary — what the Agent can sense. The input side: all pathways through which information reaches the Agent from outside itself.
- Cognitive Boundary — what the Agent can model. The middle layer: all the modeling the Agent can do on the information it receives.
- Causal Boundary — what the Agent can affect. The output side: all the ways the Agent can change its surroundings.
Sub-forms and Instance Terms
The Boundaries have sub-forms that carry diagnostic value.
- Actual Input — the specific information crossing the Computational Boundary at a given position and time.
- Actual Output — the specific effect flowing through the Causal Boundary at a given position and time.
- Intrinsic Causal Boundary — the part of the Causal Boundary the Agent carries independent of position. What travels with you.
- Positional Causal Boundary — the part of the Causal Boundary the Agent has because of where it sits. What the position confers.
- Tool — an entity an Agent uses to extend one or more of its Boundaries. The contrastive category to Agent.
Perspectives and Modulators
Every Boundary can be read from two perspectives, and every Agent carries a self-model.
- Frame — the Agent’s self-model of its own Boundaries. What the Agent thinks it can sense, model, and affect.
- Capacity — the outermost Boundary itself. The maximum reach, whether or not it is currently being put to work.
- Realized — what the Agent is actually putting to work. Can touch Capacity but never exceed it.
Composition and Application
When Agents combine, the framework’s central research question appears.
- Composition — what happens when Agents combine. The result is not additive.
- Target Surface — the set of outcomes an Agent wants to see in the world.
- Trapped Intelligence — any gap between what an Agent or Composition has and what it can access or use.
Structural Concepts and Principles
These concepts appear throughout the definitions. Each is a structural claim or analytical convention the framework relies on.
- Franz Ferdinand Effect — when an Agent’s Causal reach exceeds its Cognitive modeling, producing consequences nobody predicted.
- Turtles All The Way Down — the same three-Boundary analysis applies at every scale, recursively.
- Capacity is not Value — closing the gap between Capacity and Realized is not automatically desirable.
- Frozen-in-Ice — the analytical convention that Boundaries are treated as time-and-context-free for analysis.
- Causal-to-Computational Interface — how Agents couple. One Agent’s Causal output becomes another’s Computational input.
- Topology — the structural arrangement of Agents within a Composition. Who is connected to whom.
- Protocol — the rules governing how Agents interact along the edges of a Topology.