A possibility can fail because of conditions that later stop being true. Remembering that possibility is useful, but memory alone is not enough. A system also needs a rule for deciding when changed circumstances justify returning attention to it.
Causal Reactivation is that rule. It links dormant possibilities to the conditions that currently suppress them, then uses meaningful changes in those conditions as signals for renewed evaluation.
Reactivation should have a reason
Suppose an action is currently blocked by cost, missing capability, unavailable infrastructure, insufficient evidence, timing, or a dependency. If one of those governing conditions changes, the action’s previous evaluation may no longer describe the current world.
The important part is the connection. A lower cost can matter to a cost-blocked branch. A new capability can matter to a capability-blocked branch. Neither should automatically wake possibilities whose viability is unrelated to those changes.
Why periodic retry is weaker
A naive system can periodically reconsider every rejected option. That prevents permanent forgetting, but it spends attention repeatedly asking questions whose answers have no reason to be different.
Causal Reactivation makes reconsideration selective. Instead of asking whether enough time has passed, the system asks whether something relevant to the old judgment has changed.
The three-part memory chain
Conditional Failure Memory preserves why a branch failed under a particular state. Graded Dormancy keeps that branch available without demanding full active attention. Causal Reactivation determines when evidence from the world should move it back toward evaluation.
Together, the mechanisms turn failure from a permanent label into a state-dependent record with a path back into consideration.
Reactivation can be graded too
A changed condition does not have to make a possibility immediately competitive. It may only weaken a blocker. That can justify moving the branch from deep dormancy to light dormancy, increasing monitoring, or scheduling a fresh evaluation without committing resources to execution.
This matters because the world rarely changes in one clean jump. Viability can return incrementally.
Why this matters for adaptive systems
Agents and decision systems operate inside environments that change faster than static conclusions do. Capabilities improve. Prices fall. information arrives. Dependencies become available. Constraints disappear. A system that cannot connect those changes to its stored possibilities remains governed by an obsolete map.
Causal Reactivation lets the map update selectively. The system does not need to forget what it learned, and it does not need to reopen every closed branch. It needs to know which old conclusions were conditional on facts that are no longer the same.