Dissociation as Simulation Bypass: Acting on Unverified Planning
by Nick Clark | Published March 27, 2026
Dissociation in the cognitive architecture occurs when execution proceeds directly from speculative planning graph content, bypassing the containment boundary entirely. The agent acts on simulated scenarios as if they were verified reality. Unlike containment collapse where the boundary degrades gradually, dissociation represents a direct bypass where execution pathways route around the containment boundary rather than through it.
What It Is
Dissociation bypass occurs when the execution pathway discovers a route from speculative planning graphs directly to action execution without traversing the containment boundary. The planning graph content is never promoted through the standard validation pipeline; it reaches the execution system through an architectural shortcut that circumvents containment.
Why It Matters
Dissociation bypass is distinct from containment collapse. In collapse, the boundary degrades and everything leaks through. In dissociation, the boundary may be intact but specific execution pathways route around it. This means containment monitoring may report healthy boundary integrity while the agent is executing on unverified content through the bypass.
How It Works
The bypass forms when execution pathway routing develops a connection between planning graph outputs and execution inputs that does not traverse the containment checkpoint. This can occur through delegation chains that skip containment steps, through memory field mutations that reclassify speculative content as verified, or through execution context creation that references planning graph content directly.
Detection requires monitoring not just containment boundary integrity but also execution input provenance: tracing every executed action back to verify it originated from verified memory through the standard promotion pathway.
What It Enables
Understanding dissociation as a routing bypass rather than a boundary failure enables targeted prevention: securing all execution input pathways, not just the containment boundary. It also explains why some agents can pass containment health checks while still exhibiting dissociative behavior: the checks evaluate the boundary, but the problem is in the routing.
