Two Agent Theory

Back to Game Theory

A two-agent system is called dyadic because it contains exactly two strategic entities whose choices directly influence one another. This is the foundational structure for all classical game theory, and it introduces the first true interpersonal dynamics.

In a two-agent (dyadic) system, all relationship dynamics can be fully described by two fundamental dimensions: goal relationship and structural symmetry. Goal relationship captures whether the agents’ objectives are aligned or opposed, defining the motivational structure of the interaction. Structural symmetry captures whether the agents operate with equivalent roles, actions, and information, or with asymmetric ones, defining the interaction’s strategic structure. Together, these two dimensions generate the complete 2×2 space of dyadic game types and form the foundation for all two-agent relationship dynamics.

Goal Relationship – Structural Symmetry – Information Relationship – Commitment / Binding – Temporal Structure – Control Topology

First Dimensions of a Two-Agent Theory Scenario

(4 More Dimensions Below)

Goal Relationship – Motivation

Goal relationship describes how the objectives of two agents relate within a dyadic system. It is the motivational axis of interaction — the dimension that determines whether the agents are working with each other, against each other, or operating within a partially shared structure of incentives. Formally, two agents may have aligned goals, in which both evaluate outcomes identically and seek the same end state, or opposed goals, in which the success of one agent necessarily diminishes or prevents the success of the other. This distinction establishes the foundational motivational structure of the dyad and defines whether the interaction is cooperative, competitive, or situated somewhere between those extremes.

Structural Symmetry – Capability Footing

Structural symmetry describes whether two agents in a dyadic system operate with identical strategic capacities or with role-differentiated capacities. A symmetric interaction is one in which both agents possess the same action sets, information access, and structural roles; each could, in principle, step into the other’s position without altering the game’s formal properties. An asymmetric interaction, by contrast, is defined by non-equivalent roles, where agents differ in their available actions, information, constraints, or payoff structures. Structural symmetry therefore establishes the strategic architecture of the dyad: whether agents confront each other as equals within the same framework, or engage from fundamentally different positions within the system.

What Results from these two Dimensions:

Two-Agent Theory Scenarios

Fully Cooperative (Aligned Goals + Symmetric Structure)

In this configuration, both agents pursue the same objective and evaluate outcomes identically, and they do so from structurally equivalent positions. Each agent has access to the same types of actions, information, and strategic possibilities, allowing them to function as interchangeable partners within the system. Because their goals are aligned and their roles are symmetric, the interaction is defined by joint optimization: the success of one agent is inseparable from the success of the other. Strategic behavior therefore centers on coordination, division of labor, timing, and mutual reinforcement rather than opposition or negotiation. This cell represents the purest form of cooperation possible in a dyadic game.

Examples:

Agents share both the goal and the role.

Role-Differentiated Cooperation (Aligned Goals + Asymmetric Structure)

In this configuration, both agents pursue a shared objective, but they operate from structurally unequal roles within the system. Their goals are aligned, yet each agent has different actions, information, or capabilities, creating cooperation through specialization rather than interchangeability. Strategic effectiveness depends on how well these differentiated roles coordinate toward the common aim, with each agent contributing according to its unique functional position. This category captures cooperative dynamics grounded in complementary roles—such as coordinated hunts with defined responsibilities, paired technical tasks, or any dyadic system where collaboration is essential but the agents do not occupy identical strategic footing.

Examples:

Agents share the goal but not the role.

Strictly Competitive (Opposed Goals + Symmetric Structure)

In this configuration, the two agents pursue mutually exclusive objectives while operating from structurally identical positions. Each agent has access to the same actions, information, and strategic possibilities, and each evaluates outcomes as the exact inverse of the other. Because the roles are symmetric and the goals are fully opposed, the interaction reduces to pure contest: every gain for one agent is an equivalent loss for the other. This category encompasses the classical zero-sum structure of dyadic conflict, where strategic behavior is centered on direct opposition, mirrored capabilities, and the continual attempt to outmaneuver an identically empowered adversary.

Examples:

Agents oppose each other from the same structural position.

Asymmetric Conflict (Opposed Goals + Asymmetric Structure)

In this configuration, the two agents pursue opposed objectives while occupying structurally unequal roles within the system. Their goals are antagonistic, but the strategic landscape is not mirrored: each agent possesses different actions, information sets, constraints, or capabilities. Because the roles are asymmetric, the conflict does not unfold as a balanced contest but as an interaction shaped by differentiated strengths and vulnerabilities. Strategic behavior emerges from how each agent exploits its own role-specific advantages while compensating for its inherent limitations. This category captures antagonistic dynamics such as predator–prey pursuit, attacker–defender engagements, evasion-and-capture scenarios, and any dyadic conflict defined by unequal strategic footing.

Examples:

Agents oppose each other from different structural positions.

Additional Dimensions of Two Agent Theory

(Will be addressed within the Scenarios)

Information Relationship – Knowledge

The information relationship defines what each agent knows, what each agent does not know, and how that knowledge is distributed between them. Even when goals and structural capabilities are fixed, differences in information fundamentally alter behavior by shaping prediction, signaling, deception, learning, and coordination. Two agents with identical incentives and powers will act differently depending on whether relevant state is visible or hidden and whether knowledge is shared or unevenly held. This dimension governs uncertainty, surprise, and informational leverage.

Commitment / Binding – Reversibility

Commitment and binding describe the degree to which choices constrain future action. This dimension determines whether intentions are reversible, whether promises are credible, and whether agents can defect after signaling or acting. Binding can arise internally (self-commitment), mutually (agreements), or externally (rules, force, institutions). Holding all else equal, interactions differ radically depending on whether agents are free to change course or locked into prior decisions. This dimension governs stability, trust, and credibility.

Temporal Structure – Future Linkage

Temporal structure specifies how interaction unfolds over time. Whether an interaction is one-shot, finitely repeated, indefinitely repeated, or continuous determines whether memory, reputation, retaliation, and learning are possible. Time is not an accessory to strategy; it is the medium in which strategy exists. Identical interactions produce different outcomes depending on horizon and sequencing. This dimension governs anticipation, patience, escalation, and long-run equilibrium behavior.

Control Topology – Causal Wiring

Control topology defines how agents’ actions causally combine to produce outcomes. It specifies whether decisions are made simultaneously or sequentially, whether one agent leads or both must consent, and whether control is shared, vetoed, or unilateral. Even with identical goals, information, commitment, and timing, outcomes can diverge purely due to how control is wired. This dimension governs coordination success, dominance, deadlock, and failure modes arising from structure rather than intent.