Goal relationship = whether agents evaluate outcomes in the same direction or in opposing directions.
Goal relationship defines how the objectives of two agents relate to one another. It specifies whether the agents evaluate outcomes in the same direction, in opposing directions, or independently.
It answers the most basic structural question in any interaction:
Does one agent’s success help, harm, or leave unchanged the other agent’s success?
This relationship establishes the motivational geometry of the interaction. Before strategy, before information, before commitment or time, goal relationship determines whether agents are structurally incentivized to support one another, resist one another, or remain neutral. All higher-level dynamics—coordination, conflict, bargaining, deception, enforcement—operate within the constraints set by this relationship.
Everything else—strategy selection, information use, commitment mechanisms, temporal incentives, and control structure—rests on top of this foundation. If goal relationship is misidentified, downstream analysis becomes incoherent: cooperation is mistaken for altruism, conflict for hostility, and structural incentives are replaced with narrative explanations. Correctly specifying goal relationship is therefore a prerequisite for any valid analysis of two-agent interaction.
What a “Goal” Is (Precise Definition)
A goal is the criterion an agent uses to evaluate and rank outcomes. It is the ordering that determines which states of the world are better, worse, or equivalent from the agent’s perspective.
A goal is not:
- emotions or feelings
- stated intentions or plans
- narratives or justifications
- declared statements or promises
These may accompany action, but they do not define what an agent is optimizing.
A goal is:
- the pattern of outcomes the agent consistently prefers over others
- the objective function the agent is effectively optimizing, whether explicit or implicit
- the standard by which success and failure are judged in practice, not in rhetoric
Goals are inferred from revealed preference, not from language.
Goal relationship, therefore, concerns preference alignment, not behavior. Two agents may behave cooperatively while holding opposed goals, or behave antagonistically while holding aligned goals, depending on information, constraints, or time. The goal relationship is determined solely by how each agent evaluates outcomes, independent of how they act in any particular moment.
The Core Split (There Are Only Two)
At the dyadic level, goals can relate in exactly two structurally distinct ways:
- Aligned
- Opposed
This is not a stylistic choice or a modeling convenience. It is an exhaustive structural split. Any apparent third category reduces to one of the two once ambiguity is removed.
There is no stable third goal relationship that is not:
- a mixture of multiple goals collapsed into one description,
- a product of uncertainty about true objectives, or
- an aggregation artifact arising from time, population, or scope.
Once goals are defined as outcome-evaluation criteria, their relationship must either point in the same direction or in opposing directions. All dyadic interactions instantiate one of these two cases. Anything that appears intermediate belongs to a different dimension of analysis, not a third goal relationship.
1) Aligned Goals
Two agents have aligned goals when they evaluate outcomes using the same ordering of preference.
This does not mean:
- the agents benefit equally
- the agents occupy identical roles
- the agents agree on methods or tactics
- the agents behave kindly, cooperatively, or politely
Alignment is not about symmetry, behavior, or tone. It is about how outcomes are ranked.
It does mean:
If outcome A is preferred to outcome B by one agent, then outcome A is also preferred to outcome B by the other.
The agents may disagree about execution, timing, or responsibility, but they do not disagree about which outcomes are better or worse in principle.
Structural implications
When goals are aligned:
- There is no inherent incentive to sabotage the other agent.
- Failure arises from coordination breakdown, not betrayal.
- Disagreement concerns how to achieve the objective, not what objective should be achieved.
Any action that appears obstructive must be explained by information gaps, miscoordination, role asymmetry, or control structure—not by opposing incentives.
Important clarifications
Aligned goals can still involve:
- Tradeoffs — one agent may bear more cost than another.
- Sacrifice — one agent may accept loss for the shared objective.
- Hierarchy — authority and decision power may be unequal.
None of these introduce goal opposition. They modify structure, not preference alignment.
Boundary condition
The defining feature of aligned goals is the absence of fundamental incentive opposition. There is no outcome in which improving one agent’s position makes the other worse off by that agent’s own evaluative criteria.
2) Opposed Goals
Two agents have opposed goals when there exists at least one outcome that is preferred by one agent and dispreferred by the other.
This does not require:
- hatred or hostility
- adversarial intent
- zero-sum payoffs
- physical conflict or violence
Opposition is not psychological or moral. It is structural.
It does mean:
There exists at least one outcome in which improving one agent’s position worsens the other agent’s position, according to their own outcome-evaluation criteria.
The agents may agree on many outcomes, share partial interests, or cooperate tactically, but their preference orderings are not fully aligned.
Structural implications
When goals are opposed:
- Incentives are inherently in tension.
- Strategic resistance is rational, not pathological.
- Any sustained cooperation must be explained by temporal structure, commitment mechanisms, or external enforcement.
Absent such stabilizing factors, agents are structurally motivated to block, counter, or constrain one another when interests diverge.
Important clarifications
Opposed goals do not imply:
- that every interaction is adversarial
- that gains must be equal and opposite
- that conflict will always manifest
Opposition creates the possibility of conflict, not its inevitability. Whether conflict occurs depends on information, time horizon, binding constraints, and control topology.
Boundary condition
The defining feature of opposed goals is the presence of at least one outcome where one agent’s improvement constitutes a loss to the other by that other agent’s own evaluative standard. Once such an outcome exists, the interaction is structurally non-aligned, regardless of surface cooperation or civility.
What Goal Relationship Is Not
This is the point at which analysis most often fails.
Goal relationship is not:
- Payoff magnitude (positive-sum vs zero-sum)
- Moral framing (good vs bad, right vs wrong)
- Emotional stance (friendly vs hostile)
- Temporary tactics (“we’re cooperating right now”)
- Outcome uncertainty (not knowing what will happen)
These concepts describe outcomes, attitudes, or states of knowledge. They do not describe how agents evaluate outcomes.
Goal relationship concerns preference structure, not surface behavior or narrative interpretation.
You can have:
- Opposed goals with positive-sum outcomes
Both agents gain relative to baseline, but one still prefers some outcomes the other rejects. - Aligned goals with unequal rewards
Both rank outcomes identically even if benefits are distributed unevenly. - Opposed goals with peaceful behavior
Conflict is possible but suppressed by time horizon, enforcement, or mutual constraint. - Aligned goals with intense disagreement
Disputes concern execution, authority, or sequencing—not desired outcomes.
None of these combinations alter the underlying goal relationship.
Boundary rule
If two agents disagree about which outcomes are better or worse in principle, their goals are opposed.
If they do not, their goals are aligned—regardless of how cooperative, adversarial, polite, or chaotic their behavior appears.
Mixed, Partial, or Changing Goals
You will often hear claims such as:
- “Goals are partially aligned.”
- “Goals are mixed.”
- “Sometimes aligned, sometimes opposed.”
Taken at face value, these suggest a third category. There is none.
Structurally, all such statements reduce to one of three resolvable cases:
1) Aggregation Artifact
Multiple distinct goals have been collapsed into a single description.
- Different objectives are bundled together
- Alignment differs across those objectives
- The apparent “mix” disappears once goals are decomposed
Correction: disaggregate the goals and evaluate alignment per objective.
2) Temporal Shift
Goals are aligned at one point in time and opposed at another.
- Alignment changes across phases
- Incentives evolve as conditions change
- Conflict is time-indexed, not intrinsic
Classification: this is a temporal structure issue, not a new goal type.
3) Information Uncertainty
The true goals of one or both agents are not fully known.
- Declared goals differ from revealed preferences
- Intentions are inferred, not observed
- Apparent alignment reflects belief, not reality
Classification: this is an information relationship issue, not a goal relationship.
Boundary conclusion
Once aggregation is resolved, time is indexed, and uncertainty is removed, the goal relationship is always either aligned or opposed. There is no stable third static category.
Any claim of “mixed” goals indicates a modeling error in scope, time, or information—not a new form of preference relationship.