This dimension determines whether system execution is mediated through an explicit shared world state or through agent-local and developer-managed state.
Graph-based and role-based frameworks do not maintain a first-class environment. They rely instead on workflows and task context effects that are not reflected in a persistent global state.
GABM frameworks centralize execution around an explicit environment that tracks world state and applies agent actions over time. This environment maintains grounded variables that represent the current state of the simulated world.
| Framework | Environment |
|---|---|
| Graph-based | None; execution context only |
| Role-based | None; task-centric context |
| GABM | Explicit world state with grounded variables |
In GABM frameworks, the environment maintains a persistent world state that evolves over time. This state consists of grounded variables that represent the current condition of the simulated system. Agents observe this state, make decisions, and their actions update the environment through transition functions.
Graph-based and role-based frameworks operate without such an explicit environment. They manage state through execution context, workflow variables, or task-specific memory, but these do not constitute a persistent, shared world state that all agents can observe and modify.
Trade-off: Explicit environments enable persistent state evolution and simulation-based analysis, but add execution overhead and reduce direct control. Workflow-based execution is more efficient but lacks persistent global state.