Control topology = the wiring of agency: the mapping from each agent’s action to the resulting state/outcome, including order, veto, and override.
Control topology describes the causal wiring of agency: the way two agents’ decisions are structured to combine into outcomes. It specifies not what agents want or know, but how their choices are permitted, sequenced, constrained, and resolved within the system.
It answers a single structural question:
Who is allowed to decide what, when, and with what form of access to the system’s levers of action?
Even when agents share identical goals, possess identical information, operate under identical commitments, and face the same time horizon, outcomes can still diverge purely because the control topology differs. This is the dimension where coordination succeeds or fails, where deadlock emerges despite agreement, and where dominance can exist without superior intent or information.
Control topology can be understood as the combined structure of decision rights, move order, veto points, and coupling—the formal rules that determine how individual choices propagate through the system to produce collective results.
What Control Topology Includes (and What It Excludes)
Control topology specifies the formal mechanics by which agent choices are transformed into outcomes. It is concerned solely with the structure of decision authority, sequencing, and resolution—independent of motivation, knowledge, or time horizon.
Includes
- Move structure
Whether agents act simultaneously or sequentially, and whether any agent observes another’s action before choosing their own. - Authority structure
How decision rights are distributed: unilateral, shared, or delegated, including whether one agent sets constraints within which the other operates. - Veto structure
Which agents can block actions or outcomes, under what conditions, and whether consent is required for change. - Coupling
Whether one agent’s action directly affects the other’s outcome, or only indirectly through changes in shared state. - Update mechanism
How the system resolves competing or combined actions: by averaging, priority, threshold, override, or other formal resolution rules.
Excludes
- What agents want
Goals and incentives belong to goal relationship, not control topology. - What agents know
Information availability and distribution belong to information relationship. - Whether choices are locked or reversible
Binding and enforcement belong to commitment / binding. - Whether interaction has a future
Repetition, horizon, and persistence belong to temporal structure.
Control topology is strictly concerned with one thing:
how choices become reality.
If you change who can act, who can block, who moves first, or how actions are resolved, you have changed the control topology—even if everything else remains identical.
The Canonical Control Topologies (Minimal Set)
The control topologies that follow are not illustrative examples or situational variants. They are structurally distinct wiring types—the minimal set of ways in which decision authority, sequencing, and resolution can be arranged between two agents. Each topology defines a different causal mapping from individual choices to outcomes, independent of goals, information, commitment, or time horizon.
Together, these topologies exhaust the space of how agency can be formally connected in a dyadic system. Any two-agent interaction can be represented as one of these topologies or as a direct composition of them. The differences between them explain why identical agents, facing identical incentives, can produce deadlock, dominance, coordination, or failure purely as a result of structural design.
The sections below expound each topology in turn, detailing its defining structure, strategic implications, and characteristic failure modes.
1) Simultaneous Independent Control
In simultaneous independent control, both agents select their actions without observing the other’s choice, and the outcome is determined by the joint combination of those actions. Neither agent has temporal priority or informational access to the other’s move at the moment of decision. Strategic interaction therefore occurs entirely through anticipation, not reaction.
This topology treats agency as parallel and uncoupled at the moment of choice, even though outcomes may be tightly coupled after the fact. Any dependence between agents arises from how their actions interact in the system, not from control over timing or sequencing.
Signature
- No action observability at decision time
Each agent must act without knowing what the other will do in that moment. - Strategic uncertainty arises from prediction
Success depends on accurately anticipating the other agent’s choice rather than responding to it. - No first-mover advantage by construction
Neither agent can condition their action on the other’s current move.
Produces
- Coordination problems under aligned goals
When agents want the same outcome, success requires selecting compatible actions without confirmation, creating risk of mismatch despite shared intent. - Mixed strategies and bluffing pressure under opposed goals
When goals are opposed, agents may randomize or disguise intent to avoid being exploited through predictability. - Symmetry of strategic exposure
Because neither agent can react in real time, both face equivalent strategic risk at the point of decision.
Common Failure Mode
- Miscoordination due to simultaneous mismatch
Failure occurs when agents choose incompatible actions—not because of disagreement, deception, or incompetence, but because simultaneous choice prevents alignment at the moment it is required.
This failure mode is structural: it persists even when agents are rational, informed, cooperative, and well-intentioned.
2) Sequential Control
In sequential control, agents do not act in parallel. One agent chooses first, and the other responds after observing the first agent’s action, either fully or partially. This establishes a temporal ordering that directly affects strategic power, even when all other factors—goals, information, commitment, and horizon—are held constant.
The defining feature of this topology is that timing itself becomes a strategic resource. The first move shapes the decision context of the second, while the second move derives its effectiveness from what can be observed and inferred about the first.
Signature
- A defined move order exists
One agent acts before the other, creating asymmetry in temporal access even if roles are otherwise symmetric. - Observability conditions matter
The second agent’s ability to respond depends on whether the first move is fully visible, partially visible, or noisy. - Reactive power is conditional
The second agent’s influence is constrained by what the first agent has already committed to or revealed.
Produces
- First-mover advantage or disadvantage, depending on context
Acting first can confer leverage by shaping the option space, or it can create vulnerability by exposing intent. - Commitment effects
The initial move can function as a binding or semi-binding commitment, intentionally or unintentionally restricting future options. - Strategic signaling through action
Because the first move is observed, it may be used to signal strength, intent, or type—truthfully or deceptively.
Common Failure Modes
- Exploitation of predictable response
If the second agent’s reactions are anticipated, the first agent can manipulate outcomes by preemptively steering choices. - Over-commitment by the first mover
Acting early can lock the first agent into a disadvantageous position if the response space is misjudged.
These failures arise not from poor incentives or lack of information, but from the structural consequences of move ordering itself.
3) Leader–Follower (Stacked / Command Topology)
In leader–follower control, one agent’s decision does not merely precede the other’s—it defines the decision space itself. The leader sets policies, constraints, or parameters that shape what actions are available, while the follower optimizes their behavior within the frame established by the leader. This topology represents structural role control, not just temporal ordering.
Unlike simple sequential control, the asymmetry here is persistent and role-based. The follower’s agency is real but bounded: effectiveness depends on how well the leader’s framing aligns with the system’s demands and the follower’s incentives.
Signature
- Structural role differentiation
One agent occupies a framing role (policy, constraint-setting), while the other occupies an execution or adjustment role. - Constrained action space for the follower
The follower’s choices are shaped—sometimes tightly—by the leader’s prior decisions. - Authority persists across actions
The leader’s control is not exhausted by a single move; it recurs as part of the system’s design.
Produces
- Efficient coordination under aligned goals
When objectives are shared, this topology reduces ambiguity and accelerates coordination by centralizing framing decisions. - Systematic dominance under opposed goals
When goals conflict, the leader’s control over the action space creates enduring strategic advantage rather than momentary leverage. - Reduced need for real-time negotiation
Many decisions are resolved implicitly by the structure of authority rather than explicit agreement.
Common Failure Modes
- Bottlenecks and fragility
Over-centralization can make the system brittle, with performance and resilience tied to the leader’s judgment and capacity. - Dependence on leader competence
Errors or blind spots at the framing level propagate downstream, often magnified by execution. - Strategic sandbagging by the follower
If incentives are misaligned, the follower may comply minimally, delay, or underperform while remaining formally within constraints.
These failure modes are structural. They persist even when both agents are informed, rational, and acting in good faith.
4) Mutual Consent / Veto (Joint Control)
In mutual consent control, certain actions or outcomes can occur only if both agents approve. Each agent possesses veto power, meaning either can block change by withholding consent. The defining feature of this topology is that inaction is the default: unless agreement is reached, the system remains in its current state.
This topology reallocates power from the ability to act to the ability to prevent action. Influence is exercised not by initiating moves, but by controlling whether moves are allowed to take effect.
Signature
- A veto point exists
Each agent has the formal ability to block specified actions or outcomes. - Default-state persistence
The system remains unchanged unless both agents (or a defined quorum) agree to alter it. - Symmetric blocking power
Neither agent can force change unilaterally; control is exercised through consent.
Produces
- Stability and safety under aligned goals
When objectives are shared, veto power prevents unilateral mistakes and protects against rash or misaligned actions. - Negotiation-centered dynamics
Progress requires explicit agreement, making bargaining, compromise, and signaling central features of interaction. - Hostage dynamics under opposed goals
When incentives conflict, the veto becomes a leverage tool, shifting power to whoever can most afford delay or stalemate.
Common Failure Modes
- Negotiation gridlock
Mutually beneficial outcomes may fail to occur because agreement cannot be reached, even when no agent prefers the status quo. - Concession extraction via veto leverage
One agent may withhold consent to force side payments, policy changes, or asymmetric gains unrelated to the blocked action itself.
These failures are not pathologies of cooperation or bad faith; they are structural consequences of shared blocking authority.
5) Unilateral Override (Dominant Control)
In unilateral override control, one agent’s choice determines the outcome regardless of the other agent’s action. The dominant agent possesses decisive control authority: their decisions either override, nullify, or render irrelevant the actions of the second agent. The weaker agent retains agency only in a reactive or indirect sense, unless they can alter the control topology itself.
This topology represents maximal asymmetry of control. Strategic interaction no longer centers on coordination or negotiation, but on enforcement, compliance, and resistance.
Signature
- Override authority exists
One agent’s actions supersede the other’s, either by direct command, rule enforcement, or structural priority. - Nullification of subordinate action
The second agent’s choices do not directly affect outcomes unless they interfere with enforcement or system stability. - Topology-dependent agency
The weaker agent’s strategic relevance depends entirely on their ability to evade, undermine, or transform the control structure.
Produces
- Predictable outcomes for the dominant agent
Results are stable and foreseeable so long as enforcement holds and control remains intact. - Compliance-driven coordination
Alignment, if present, is enforced rather than negotiated; cooperation becomes procedural rather than voluntary. - Resistance-oriented strategy for the weaker agent
Strategic effort shifts away from direct contest toward evasion, sabotage, delay, or attempts to change the topology itself.
Common Failure Modes
- Collapse of cooperation
The weaker agent disengages, withdraws effort, or ceases meaningful participation when their actions no longer influence outcomes. - Instability under undermined enforcement
If the dominant agent cannot reliably enforce override authority, resistance escalates and control rapidly erodes.
These failures are structural. They arise not from misaligned goals or poor information, but from the inherent brittleness of dominance when enforcement is costly or incomplete.
Two Additional Properties That Matter Inside Any Topology
The properties below do not define new control topologies. They are modifiers that operate within any topology, shaping behavior, stability, and failure modes without changing who formally controls decisions. Two systems may share the same control topology and still behave very differently depending on these properties.
A) Coupling Strength
Coupling strength describes how strongly one agent’s actions propagate through the system and affect the other agent’s outcomes.
- Tightly coupled systems
Small actions produce large downstream effects. Feedback is fast, sensitivity is high, and errors can cascade rapidly. Coordination demands precision, and failures tend to be abrupt rather than gradual. - Loosely coupled systems
Actions are mostly local in effect. Errors are dampened, feedback is slower, and disturbances are contained. Coordination is more forgiving, and failures tend to degrade performance rather than collapse it.
Structural implication:
Tight coupling amplifies both success and failure; loose coupling trades efficiency for robustness. Many brittle systems fail not because of poor incentives or bad intent, but because coupling strength exceeds agents’ capacity to coordinate.
B) Control Granularity
Control granularity describes the resolution at which agents can influence outcomes.
- Continuous control
Agents can make fine-grained, incremental adjustments. Errors can be corrected gradually, and strategies can adapt smoothly to changing conditions. - Discrete control
Decisions occur at thresholds, switches, or binary points—on/off, allow/block, escalate/stand down. Outcomes change abruptly once a boundary is crossed.
Structural implication:
Discrete control increases decisiveness but raises the risk of cliff-edge failures. Continuous control supports stability and adaptation but may slow response or obscure accountability.
Why These Properties Matter
Coupling strength and control granularity explain why systems with identical goals, information, commitment, time horizon, and control topology can exhibit radically different behavior. They determine whether a system is robust or brittle, whether errors are absorbed or amplified, and whether coordination failures are recoverable or catastrophic.
They do not alter who decides—but they strongly affect how dangerous, resilient, or forgiving those decisions are once made.
Why This Dimension Is Logically Necessary
Control topology is not an optional refinement. If it is left unspecified, the interaction itself is underdetermined. Without an explicit control topology, you literally cannot answer basic structural questions such as:
- Can either agent act unilaterally?
- Can one agent block the other’s action?
- Does timing affect outcomes?
- Does acting first confer advantage or impose risk?
- Is coordination required for change, or merely helpful?
Each of these questions determines how decisions translate into outcomes. Without answers to them, predictions about behavior, stability, or failure are meaningless. Two interactions that appear identical in goals, information, commitment, and time can behave in entirely different ways purely because these control questions resolve differently.
For this reason, control topology is not descriptive decoration. It is a required component of the system’s specification. Omitting it leaves the model incomplete.
The Boundary Rule (What It Is, Precisely)
Control topology is the set of formal rules assigning decision rights and causal precedence within the interaction.
It specifies:
- Who is permitted to initiate action
- Who can block or veto outcomes
- Who moves first or last
- Whose decisions override others
- How multiple actions are resolved into a single outcome
The boundary is strict. If you change who can veto, who moves first, or who can override, you have changed the control topology—and therefore changed the interaction itself—even if goals, information, commitment, and time horizon remain unchanged.
Control topology answers only one question, but it answers it completely:
How do choices become reality?
That question cannot be answered anywhere else in the model.