Information relationship = the pattern of observability and knowledge distribution that determines what each agent can infer, signal, conceal, and reliably predict.
Information relationship describes the structure of who knows what in a two-agent interaction, and what each agent knows about the other’s knowledge. It is not communication, skill, or cognition. It is the objective distribution of decision-relevant facts within the system.
Information includes facts about the environment, the agents, and the rules linking actions to outcomes. Crucially, it also includes higher-order knowledge—what each agent knows about what the other knows—because mutual awareness directly shapes feasible action.
Once goals (what agents want) and capabilities (what agents can do) are fixed, information relationship becomes the primary determinant of behavior. Agents with identical incentives and powers will act differently solely because they operate under different informational conditions.
Information relationship governs the following core dynamics:
- Prediction – can I anticipate you?
- Whether an agent can reliably anticipate the other’s actions, reactions, or future states of the system. Prediction depends on observability, inference quality, and belief alignment.
- Coordination – can we line up actions?
- Whether agents can align actions in time and form to achieve joint outcomes. Coordination requires not only shared intent but sufficient common knowledge to act without misalignment or hesitation.
- Deception and Concealment – can I mislead you or hide?
- Whether an agent can hide relevant state, misrepresent intentions, or manipulate the other’s beliefs. Deception is only possible where information is imperfect or asymmetrically distributed.
- Learning – do actions reveal hidden state?
- Whether actions, signals, or outcomes reveal hidden information over time. Learning determines how uncertainty evolves and whether informational advantages persist or decay.
- Power via Knowledge – do I have an informational edge?
- Whether one agent possesses an informational advantage that can be leveraged independently of formal capability. Informational power arises when knowledge substitutes for force, resources, or authority.
In short, information relationship defines the uncertainty landscape of the interaction: what is visible, what is hidden, what is shared, and what is exploitable. It determines not what agents want to do or can do, but what they can reasonably expect, credibly threaten, safely coordinate, or successfully conceal.
Core Axes of the Information Relationship
Information relationship is structured by two independent axes. Together, they fully determine the informational conditions under which a two-agent interaction operates.
Axis 1: Visibility of Relevant State
This axis describes what aspects of the world and the interaction are observable at the time decisions are made. It concerns the objective availability of decision-relevant facts, not the agents’ reasoning ability.
Perfect Information
All state that is relevant to decision-making is observable when choices are taken. Agents can see the current configuration of the world, prior actions, and any variables that affect outcomes. No strategically meaningful facts are hidden.
Perfect information eliminates uncertainty about the present state of the system. It does not eliminate complexity, difficulty, or strategic depth. It only guarantees that nothing important is concealed.
Imperfect Information
Some decision-relevant state is hidden, delayed, noisy, or otherwise unobservable. Agents must act without full knowledge of the current situation and therefore rely on inference, estimation, or assumptions.
Imperfect information introduces intrinsic uncertainty. Outcomes depend not only on actions but on how agents manage unknowns and interpret incomplete signals.
Key clarification:
“Perfect” does not mean “easy.” It means nothing relevant is hidden. An interaction can be extremely complex and still involve perfect information.
Axis 2: Distribution of Knowledge Between Agents
This axis describes how available information is distributed across agents. It concerns relative informational position, not absolute certainty.
Symmetric Information
Each agent has access to the same relevant information, or to signals that are equally informative. Neither agent holds a systematic informational advantage over the other.
Symmetry does not require completeness. Both agents may face uncertainty, but they face it on equal terms.
Asymmetric Information
One agent possesses more, better, earlier, or qualitatively different information than the other. This asymmetry creates leverage independent of formal capabilities or goals.
Informational asymmetry enables advantages such as anticipation, manipulation, screening, or concealment, even when agents otherwise appear evenly matched.
Key clarification:
Symmetric information can still be imperfect. Both agents can be equally uncertain or equally ignorant.
Why These Axes Are Independent
An interaction may be:
- Perfect but asymmetric (the state is observable, but only to one agent),
- Imperfect but symmetric (both agents face the same uncertainty),
- Imperfect and asymmetric (hidden state plus uneven access),
- Perfect and symmetric (nothing hidden, nothing unequal).
These two axes cannot be collapsed without loss of explanatory power. Together, they define the full informational landscape within which prediction, coordination, deception, learning, and informational power operate.
The Four Canonical Information Cases
The two information axes—visibility of relevant state and distribution of knowledge between agents—combine to produce four and only four distinct informational regimes. These regimes are not stylistic variations or descriptive conveniences; they are the complete and exhaustive partition of informational structure in a two-agent system.
Each case represents a fundamentally different uncertainty environment. It determines what agents can reasonably predict, what they must infer, whether deception is possible, whether learning can occur, and whether informational advantage can substitute for capability or control. Crucially, these differences persist even when goals and structural roles are held constant.
The four cases are mutually exclusive and collectively exhaustive. Any two-agent interaction must fall into exactly one of them at any given moment. If an interaction appears to exhibit features of multiple cases, it is either transitioning between regimes or has been misclassified.
The sections that follow examine each case in full. For each, the analysis identifies:
- the informational conditions that define the regime,
- the classes of behavior it enables and forbids,
- the dominant strategic dynamics it produces, and
- the characteristic failure modes that arise within it.
Only after this stratification is fixed does it make sense to analyze coordination, competition, deception, learning, or power in a disciplined way.
1) Perfect + Symmetric Information
Both agents observe all decision-relevant state. Both agents know that both agents observe it.
Nothing strategically meaningful is hidden, delayed, or privately held.
The informational environment is fully transparent and evenly shared. Any uncertainty that remains arises from complexity, execution limits, or control structure—not from lack of knowledge.
Core behavioral characteristics
- Direct coordination or direct contest
Agents act without inference about hidden state. Coordination problems are explicit, and competitive behavior reduces to optimization under known conditions. - No bluffing based on hidden state
Deception through concealment is impossible because there is nothing to hide. Strategic behavior cannot rely on misinformation about the present state of the system. - Skill, planning, and execution dominate
Outcomes depend on decision quality, foresight, timing, and error management rather than informational leverage. - Common knowledge enables precise alignment
When goals are aligned, agents can coordinate precisely because there is no ambiguity about what must be done or what the other sees.
Typical failure modes
- Miscoordination due to timing or control topology
Failures arise when agents act out of sync, act simultaneously when sequencing is required, or lack a clear control mechanism—not because they misread the state. - Overreaction or underreaction due to complexity
Agents may respond incorrectly to known information because the decision space is large or cognitively demanding, not because the information is missing or distorted.
Canonical examples (aligned goals)
These are situations where cooperation is possible without inference or trust about hidden facts.
- Two people lifting a heavy object in full view of each other
- Paired climbers on a visible route with shared maps and conditions
- Rowers in the same boat with full sensory feedback
- Two surgeons performing a synchronized procedure with full visibility
- Cooperative puzzle games with mirrored roles and complete board visibility
- Firefighters coordinating entry when the layout and hazards are fully known
In all cases, success depends on synchronization, timing, and execution—not on guessing.
Canonical examples (opposed goals)
These are contests where neither side has informational advantage.
- Chess, checkers, Go, and other perfect-information board games
- Boxing, wrestling, fencing, or evenly regulated combat sports
- Arm-wrestling or tug-of-war
- Tennis, table tennis, racquet sports
- A duel between equally equipped opponents in clear conditions
- Two companies bidding in an open, fully transparent auction
Victory arises from superior strategy or execution, not surprise.
What this case structurally forbids
- Ambush based on hidden position
- Bluffing about current state
- Exploiting private knowledge
- Screening or signaling based on concealed types
- Surprise driven by informational asymmetry
If any of these appear, the interaction is not Perfect + Symmetric.
Structural summary
Perfect + Symmetric information produces the cleanest and most legible interactions. When goals align, it enables near-ideal coordination. When goals oppose, it produces fair contests governed by skill and planning. Informational power is zero; all power must come from action, timing, or structure.
2) Perfect + Asymmetric Information
All decision-relevant state is observable in principle, but not equally observable in practice.
Nothing important is hidden from the system itself, yet one agent has privileged access, earlier access, faster access, or structurally superior visibility.
The informational environment is transparent but uneven. There is no intrinsic uncertainty about the state of the world—only inequality in who can see it clearly, promptly, or at all.
Core behavioral characteristics
- Advantage through position, access, or speed
Because the state is fully knowable, informational advantage does not arise from inference or guesswork, but from vantage point. Control over observation becomes a source of power. - No bluffing about the state itself
Deception via false claims about the present state is limited or impossible. The disadvantaged agent may not see the state, but the state is not ambiguous or manipulable once revealed. - First-mover and anticipatory advantages dominate
Earlier or clearer perception allows one agent to act before the other can respond, converting informational asymmetry into positional or strategic advantage. - Strategic behavior centers on exposure and access
Conflict and coordination revolve around gaining, denying, or equalizing visibility rather than hiding state or fabricating signals.
Typical failure modes
- Action based on stale or delayed information
The disadvantaged agent reacts correctly to the wrong moment, acting on information that is accurate but no longer current. - Structural dependence on the informed agent
Coordination fails when one agent becomes a bottleneck for awareness, slowing or misdirecting joint action. - False sense of fairness
Because the system is “perfect” in principle, asymmetry is often underestimated or ignored, leading to surprise outcomes that feel unjust rather than deceptive.
Canonical examples (aligned goals)
These are cooperative settings where shared objectives exist, but awareness is uneven.
- A driver and a navigator, where only the navigator sees the full map
- A pilot and co-pilot when one has superior instrument access
- A rescue team where one member has drone or aerial visibility
- A commander receiving real-time battlefield intelligence while subordinates do not
- A surgical lead with direct imaging access coordinating assistants
- Two engineers troubleshooting a system where only one has diagnostic readouts
Success depends on whether informational asymmetry is integrated (shared, relayed, trusted) or becomes a coordination bottleneck.
Canonical examples (opposed goals)
These are contests where one side sees first or sees better, but the state itself is not ambiguous.
- Surveillance vs target in open terrain
- Sniper with clear line of sight vs exposed opponent
- High-ground advantage in visible combat
- Security monitoring systems tracking intruders
- Market participants with faster access to public price feeds
- Open-book negotiation where one side has superior analytical tooling
Victory comes from anticipation and timing, not hidden facts.
What this case structurally forbids
- Bluffing about the current state of the world
- Exploiting hidden variables that do not exist
- Ambiguity-driven deception
- Mutual uncertainty about what is true
If uncertainty or concealment drives behavior, the interaction is not Perfect + Asymmetric.
Structural summary
Perfect + Asymmetric information produces interactions where knowledge inequality replaces uncertainty. Power arises from access, position, and speed rather than secrecy or inference. When goals align, asymmetry must be actively managed to prevent dependency or delay. When goals oppose, asymmetry converts directly into leverage and initiative.
3) Imperfect + Symmetric Information
Some decision-relevant state is hidden, noisy, delayed, or unobservable, and both agents face the same uncertainty.
Neither agent has privileged access to the truth. What is unknown is unknown to both, and both know that they are operating under the same informational limits.
The informational environment is uncertain but fair. Outcomes are shaped not by informational dominance, but by how agents reason, infer, explore, and manage risk under shared uncertainty.
Core behavioral characteristics
- Inference replaces observation
Agents must act based on estimates, models, or probabilistic beliefs rather than direct knowledge of the current state. - Risk management and robustness dominate
Because uncertainty is shared, success depends on tolerance for risk, resilience to error, and strategies that perform well across possible states. - No informational leverage over the other agent
Neither side can systematically exploit ignorance, since both are equally ignorant. Advantage must come from better reasoning, better heuristics, or better adaptation. - Learning is symmetric and gradual
Information is revealed through outcomes, experimentation, or time, and both agents update beliefs in parallel.
Typical failure modes
- Paralysis due to uncertainty
Agents delay or fail to act because the unknowns prevent confident commitment. - Overconfidence in noisy signals
Both agents may overweight weak or misleading evidence, leading to synchronized but incorrect action. - Escalation driven by ambiguity
Unclear state can cause agents to interpret the same signal differently, triggering unnecessary conflict or overreaction. - Coordination breakdown despite aligned goals
Even with shared objectives, agents may diverge in interpretation, timing, or preferred risk posture.
Canonical examples (aligned goals)
These are cooperative situations where agents work together without full visibility.
- Search-and-rescue in fog, smoke, or darkness
- Two scientists investigating an unknown phenomenon
- Joint exploration of an unmapped environment
- Early-stage product development with incomplete market data
- Emergency response during unfolding natural disasters
- Two engineers diagnosing a system with intermittent failures
Success depends on shared models, disciplined inference, and adaptive coordination, not superior information.
Canonical examples (opposed goals)
These are competitive situations where neither side knows the full state.
- Poker when both players lack complete information
- Submarine warfare with mutual concealment
- Competitive bidding under uncertain valuations
- Military maneuver in fog-of-war conditions
- Legal disputes with incomplete evidence on both sides
- Business competition in volatile or opaque markets
Advantage arises from better judgment under uncertainty, not informational dominance.
What this case structurally forbids
- Persistent informational advantage
- Exploitation of private knowledge
- Dominance through surveillance or privileged access
- Deception based on concealed state known only to one side
If one agent systematically knows more, the interaction is not Imperfect + Symmetric.
Structural summary
Imperfect + Symmetric information produces interactions governed by shared uncertainty. Power comes from inference quality, adaptability, and risk discipline rather than access or secrecy. When goals align, uncertainty threatens coordination. When goals oppose, uncertainty levels the field and rewards strategic judgment rather than informational leverage.
4) Imperfect + Asymmetric Information
Some decision-relevant state is hidden or uncertain, and one agent possesses systematically better information than the other.
Uncertainty exists in the system, but it is unevenly distributed. One agent knows more, knows earlier, knows with greater precision, or knows something categorically different.
This is the most strategically rich informational regime. It is the only case in which uncertainty and inequality coexist, allowing informational advantage to be actively converted into leverage.
Core behavioral characteristics
- Inference and exploitation coexist
The less-informed agent must infer under uncertainty, while the better-informed agent can choose whether to reveal, conceal, distort, or selectively disclose information. - Deception and manipulation become structurally possible
Because some facts are hidden and unevenly held, the informed agent can shape the other’s beliefs through signaling, omission, framing, or outright deception. - Strategic behavior centers on belief management
Outcomes depend not just on actions taken, but on how beliefs are induced, maintained, or corrected over time. - Information substitutes for capability
Informational advantage can compensate for weaker formal power, resources, or position, or amplify existing asymmetries into dominance.
Typical failure modes
- Adverse selection
The less-informed agent unknowingly selects unfavorable partners, actions, or commitments because hidden qualities are systematically masked. - Moral hazard
The informed agent takes hidden risks or exploits concealed slack because consequences are not fully observable. - Misplaced trust or over-skepticism
The uninformed agent either trusts signals that should not be trusted or rejects truthful signals out of fear of manipulation. - Entrenched exploitation
Persistent asymmetry allows one agent to extract value repeatedly without triggering correction or retaliation.
Canonical examples (aligned goals)
These are cooperative settings where shared objectives exist, but knowledge is uneven.
- A senior expert guiding a novice through a complex task
- A doctor diagnosing a patient who lacks medical knowledge
- A manager with full financial data coordinating a team
- A software architect directing developers without full system context
- An intelligence analyst briefing policymakers
- A pilot flying with instruments unavailable to passengers
Success depends on responsible disclosure and trust calibration. Failure occurs when asymmetry produces dependency, blind obedience, or misalignment.
Canonical examples (opposed goals)
These are conflicts where informational imbalance is decisive.
- Poker with skilled players exploiting weaker ones
- Market trading with insider or superior analytical knowledge
- Negotiations where one side knows the true constraints or alternatives
- Espionage, counterintelligence, and covert operations
- Predator–prey dynamics with asymmetric awareness
- Legal disputes where one side controls critical evidence
Victory hinges on signal control, belief shaping, and timing, not brute force.
What this case structurally permits (uniquely)
This is the only informational regime that allows:
- Bluffing with real leverage
- Screening and signaling equilibria
- Strategic misinformation
- Sustained exploitation of ignorance
- Power without overt control
If these behaviors are present, the interaction must involve Imperfect + Asymmetric information.
Structural summary
Imperfect + Asymmetric information defines interactions where uncertainty is weaponized. Power flows from knowing what the other does not—and from deciding when, whether, and how that knowledge is revealed. When goals align, this regime demands ethical restraint and institutional safeguards. When goals oppose, it is the primary engine of manipulation, exploitation, and strategic dominance.
“Information About the World” vs “Information About the Agent”
Uncertainty in a two-agent interaction arises from two fundamentally different sources, which are often conflated but must be kept separate. One concerns the state of the world. The other concerns the nature of the other agent. These uncertainties behave differently, resolve differently, and enable different strategic dynamics.
World Uncertainty
World uncertainty concerns incomplete or noisy information about the external state of the system—facts that exist independently of the agents’ intentions or character.
This includes:
- Fog, occlusion, or limited visibility
- Hidden or partially observable state
- Noise in measurement or signals
- Randomness or stochastic processes
- Delays or gaps in observation
- Limits of sensing, instrumentation, or access
World uncertainty affects what is true right now, regardless of who the agents are. It is typically reduced through observation, measurement, exploration, or time. When resolved, it collapses to fact.
Strategically, world uncertainty:
- Forces inference rather than direct reaction
- Rewards robustness and error tolerance
- Creates risk without intent
- Applies symmetrically unless access is uneven
World uncertainty alone does not permit deception. If no agent has privileged access to hidden state, uncertainty remains environmental rather than adversarial.
Type Uncertainty (About the Other Agent)
Type uncertainty concerns incomplete knowledge about the other agent’s internal characteristics—properties that shape how they will act across situations.
This includes uncertainty about:
- Competence or skill level
- Honesty or willingness to deceive
- True objectives or hidden incentives
- Response under pressure or threat
- Risk tolerance and thresholds
- Constraints, limits, or bluffing capacity
Type uncertainty affects how the other agent will behave, not what the world currently is. It cannot be resolved by observation alone; it is inferred through behavior, history, signaling, and consistency over time.
Strategically, type uncertainty:
- Makes trust and distrust meaningful
- Enables signaling, screening, and reputation
- Persists even in stable environments
- Allows deception without hidden world state
Type uncertainty is why humans care so deeply about credibility, reputation, and “tells.” These are mechanisms for inferring agent type when direct verification is impossible or costly.
Why the Distinction Matters
World uncertainty and type uncertainty interact but are not interchangeable.
- World uncertainty is typically reduced by seeing more.
- Type uncertainty is reduced by watching behavior over time.
An interaction with perfect knowledge of the world can still be strategically unstable due to uncertainty about the agent. Conversely, agents may fully trust each other while remaining uncertain about the environment.
Failing to separate these leads to analytical errors—misattributing deception to noise, mistaking risk for malice, or treating environmental uncertainty as strategic manipulation.
Structural takeaway
World uncertainty determines what can be known.
Type uncertainty determines who can be trusted.
Both belong under Information Relationship, but they drive fundamentally different strategic phenomena.
Higher-Order Knowledge: Why “I Know That You Know” Matters
Higher-order knowledge—what each agent knows about what the other knows—is not academic abstraction. It is operational structure. Many strategic outcomes depend not on facts alone, but on whether those facts are mutually known, and whether that mutual knowledge is itself known.
Information relationship therefore includes not just raw facts, but the awareness structure surrounding those facts.
Coordination Requires Common Knowledge
Successful coordination requires more than shared information; it requires common knowledge. If both agents know the plan, but one agent is unsure whether the other knows that they know it, coordinated action may fail or be delayed.
Coordination breaks not because of disagreement, but because of uncertainty about mutual awareness. Agents hesitate, wait, or act out of sequence to avoid unilateral exposure.
Deterrence Requires Known Capability
Deterrence depends on the other agent knowing that retaliation is possible and credible. A hidden capability may exist in fact, but if it is not known—or not known to be known—threats lose force.
Deterrence fails when:
- capabilities are concealed,
- awareness is asymmetric,
- or credibility is uncertain.
Power requires not just possession of capability, but recognized possession.
Bluffing Requires Uncertainty About Type or State
Bluffing is only possible when the other agent is uncertain about relevant aspects of your type or the state of the world. If all facts and types are common knowledge, bluffing collapses immediately.
Bluffing therefore relies on:
- imperfect information,
- asymmetric beliefs,
- and limited higher-order awareness.
Remove uncertainty at the awareness level, and bluffing becomes impossible regardless of incentives.
Structural Implication
Information relationship is not just a collection of facts. It is facts plus the structure of mutual awareness surrounding those facts. Many strategic phenomena—coordination, deterrence, signaling, deception—exist or fail entirely based on this higher-order structure.
Key rule:
Facts determine what is possible.
Mutual awareness determines what is credible, coordinated, or enforceable.
Boundary of the Information Relationship Dimension
The information relationship dimension concerns knowledge structure only: what is known, what is hidden, and how that knowledge is distributed between agents. It does not describe how choices are constrained, how interaction unfolds over time, or how actions are causally combined.
Specifically, information relationship does not determine:
- Whether choices are locked in
Irreversibility, promises, contracts, and enforcement belong to Commitment / Binding, not information. An agent may know everything and still be free to defect, or know little and still be fully constrained. - Whether the interaction repeats
One-shot versus repeated interaction, horizons, and memory belong to Temporal Structure. Information may accumulate over time, but time itself is a separate structural dimension. - Whether moves are simultaneous or sequential
Turn order, first-mover advantage, veto power, and unilateral control belong to Control Topology, not information. Knowing the state does not determine how or when actions are taken.
These dimensions interact with information and often amplify or dampen its effects, but they are distinct knobs. Collapsing them into information obscures causality and leads to misclassification of strategic dynamics.
Boundary rule:
Information relationship answers what is known.
The other dimensions answer what can be done, when it can be done, and whether it can be undone.