Causal Map¶

A causal map draws attributed claims about how factors influence one another as a directed network, without treating those claims as proven effects.

Version: v2 · 2026-10-03 · History
Domain-specific #: 13048
Domain group: Social Sciences
Origin domain: Political Science
Subdomain: Cognitive Mapping → Political Science
Aliases: Cause map, Causal cognitive map

Core Idea¶

A causal map represents asserted causal influences as a directed network: nodes name factors or concepts, and arrows say that one is thought to affect another. Crucially, a map records someone's model or hypothesis, not automatically an established causal mechanism. Axelrod's original political-decision work drew concepts and causal links from decision makers' statements, then inspected the resulting graphs to understand how those actors reasoned about policy. The same structure can be elicited in workshops or coded from documents.^[1]^[2]

The abstraction separates claim structure from claim warrant. A directed edge can be traced, compared or challenged without being presumed true. Signed links, weights, path analysis, cycles and aggregation may enrich a map, but none alone defines every causal map. A map can even contain feedback cycles, so it is not necessarily a causal-inference DAG.^[1]

Structural Signature¶

Sig role-phrases:

Factors: named conditions, decisions or outcomes become nodes.
Directed links: an arrow records that a source asserts one factor influences another.
Provenance: a person, group, interview, document or workshop supplies the claim.
Coding convention: rules determine how natural-language statements become nodes and arrows.
Optional annotations: polarity, confidence, strength or temporal delay may be recorded if justified.
Interrogation: paths, cycles, contrasts between actors and candidate leverage points can be examined without claiming validation.^[1]^[2]

Condensed: source-attributed factor claims + directed graph encoding → inspectable model of asserted causal understanding.

What It Is Not¶

Not proof of causal effects. A participant's arrow is a claim to evaluate, not an experiment or identification result.^[1]
Not necessarily a DAG. Beliefs may include cycles and feedback.
Not every concept map. Generic concept maps can use noncausal links such as “is part of” or “is associated with.”
Not the same as a causal loop diagram. A CLD adds system-dynamics conventions for signed influence and reinforcing/balancing feedback.
Not automatically a quantitative structural equation model. A qualitative arrow does not supply a coefficient, functional form or identified intervention effect.
Not guaranteed consensus after aggregation. A composite may conceal disagreement among sources.

Scope of Application¶

Axelrod's original work derived graphs from policy actors' records and from judges' or experts' assessments. He explicitly distinguished accurately representing a person's causal assertions from measuring whether those assertions describe the world. Coding validity and reliability are part of the method: the graph is only as faithful as the elicitation and interpretation that produced it.^[1]

In management strategy, Eden's original description of Strategic Options Development and Analysis uses cognitive mapping with stakeholder-owned material and workshops. A team's map can reveal incompatible assumptions, possible paths from action to outcome and issues needing more evidence. Such uses do not imply that a workshop discovers one verified causal graph of the organization.^[2]

The map may support a theory-of-change discussion or provide a scaffold for a later quantitative model. That is a use of the map, not proof that its arrows are causally identified. If a project needs intervention estimates, it must add measurement, design and assumptions beyond the drawing.

Clarity¶

Each arrow should be readable as a claim with a direction, scope and source. “More training improves retention” may be one stakeholder's belief, a coded statement from a report, or an evidence-backed effect in a specified population; the map should not blur these statuses. State what the factors mean and whether a plus/minus sign refers to increasing a variable or to an evaluative judgment. A loop on paper describes a hypothesized feedback path, not measured dynamic stability.^[1]

Manages Complexity¶

A map compresses many statements into a structure that can be inspected visually and computationally. It helps people notice shared assumptions, long paths, disagreement and recurring concepts. Compression has a cost: merging synonymous factors, omitting qualifiers and turning uncertain speech into crisp arrows can manufacture clarity. Keeping source quotations, coding decisions and contested links alongside the graph limits that failure.^[1]

Abstract Reasoning¶

Choose a bounded question and identify who or what supplies the causal claims. Extract factors with a consistent granularity, encode only claims with an intelligible direction, and record ambiguities rather than force every sentence into an edge. Examine paths or loops as implications of the asserted model. Compare maps across sources before merging them. For any consequential arrow, seek independent evidence if the task requires claims about actual effects rather than beliefs.^[1]^[2]

Knowledge Transfer¶

The source–factor–arrow structure transfers among policy deliberation, management cognition and program planning. A map's interpretive usefulness does not transfer as empirical validity: graph centrality is not causal importance, path existence is not effect magnitude, and agreement among experts is not experimental proof. Distinguish the practice of representing causal beliefs from the separate methods of causal inference.

Examples¶

British Eastern Committee transcript maps¶

Axelrod derived maps of members of the 1918 British Eastern Committee from verbatim deliberation transcripts using explicit documentary coding rules. The mapped variables included policy proposals, intervening consequences and valued goals; arrows represented what members asserted about influence. Axelrod examined whether members' stated positions aligned with their own large maps. The preview attests this study design and result, not the wording of any single edge, so no particular committee arrow is invented here.^[1]

Mapped back: committee member's transcript → proposal/intermediate-outcome/goal variables → document-coded asserted causal arrows → inspectable path from an actor's stated proposal to valued outcome, without claiming the path is a measured effect.

Commuter-transport expert panel¶

Fred Roberts's transportation/energy study took a different route: a panel first assembled relevant concepts, then supplied consensus judgments about causal links after the question was narrowed to commuter transportation. Axelrod reports a tentative map-analysis result that stable intraurban commuter energy demand would require ticket prices to fall as ridership rose. This is a conclusion inside the panel-derived causal model, not an observed policy effect or an invented two-arrow transcript.^[1]

Mapped back: commuter transport and energy-demand factors → expert panel as source → questionnaire/consensus rule for directed influence claims → tentative strategic path, distinct from the Committee's actor-specific transcript maps.

Ocean-regime actor comparison¶

Hart's ocean-regime study, described in Axelrod's original volume, compared maps attributed to national and other actor groups on offshore oil, fisheries, shipping passage and related policy goals. Judges estimated actor-specific linkages and also a consensus view of the links they thought actually existed. This creates a source-critical comparison: the actor map is a judge's estimate of that actor's asserted model, while consensus is another elicited estimate, not automatic validation of either.^[1]

Mapped back: ocean-regime policy/goal factors → actor-specific influence assertions as estimated by judges → explicit questionnaire/aggregation provenance → comparison against consensus estimates while retaining their distinct warrant.

Structural Tensions¶

Graph clarity versus source fidelity. Condensing a transcript into nodes and arrows makes paths and disagreement inspectable, but can erase a speaker's hedge or the specific scope of a claim. Retaining every utterance preserves nuance but makes systematic structural comparison across a large committee difficult. Diagnostic: can an analyst trace a consequential arrow back to its speaker, wording and coding decision?^[1]

Combined account versus genuine disagreement. A panel-consensus map supports one strategic calculation, but it can suppress the fact that two actors predicted opposite consequences. Keeping only separate maps preserves who believed what, yet offers no common model for joint planning until the aggregation rule is made explicit. Diagnostic: which arrows are shared, contested or imposed by that rule?^[1]

Belief representation versus causal validation. Treating a reported arrow as a measured effect risks harmful decisions based on an untested belief. Refusing to draw or inspect any arrow until it is validated loses the map's legitimate role in surfacing testable hypotheses and actor disagreements. Diagnostic: which arrow is an attributed assertion, and what independent design or data would warrant a real-effect claim?

Structural–Framed Character¶

A causal map is mixed but strongly practice-framed on the structural–framed spectrum: nodes and directed edges are formal structure, while extracting an influence claim from speech or a workshop depends on human interpretation. Evaluative weight is high when deciding which actors, goals and disagreements matter; a map can faithfully represent a harmful or mistaken belief without endorsing it. Human elicitation and coding practice is constitutive of this attributed-belief identity, and Axelrod's political-decision program and Eden's management practice are historical institutions of use, though no one organization defines the graph. The vocabulary travels literally from elite-policy transcript analysis to participant-owned strategy mapping when source-attributed causal assertions remain the object. Importing the same diagram into causal inference does not confer measured-effect status; recognizing a directed graph in a physical system is not importing this belief-map method. Its character: a formally legible directed network whose causal-claim provenance and interpretive practice are indispensable.

Structural Core vs. Domain Accent¶

The skeletal relation is entities joined by source-attributed directed assertions. Its domain-bound mechanism is qualitative causal elicitation, coding, actor provenance and explicit separation between a person's explanation and a validated effect. The named causal map fails the prime bar because removing causal-claim semantics and source practice yields only a generic directed relation, while transferring it unchanged to a causal-inference DAG changes what its arrows warrant. The checked broader structural parent is prime Network; the edge classifies the map's node–link organization, not the truth of any causal assertion. Live Causal Loop Diagram may be a narrower child after separate semantic review, and Causal Inference is a neighboring validation practice, not automatically a parent.

This entry is a kind of Network.

Its factors and directed, asserted links have the structure of a Network, while attribution and causal semantics narrow that structure. Belonging under Network is a structural claim only: it does not validate the causal claims a map represents.

Causal Loop Diagram may be a narrower kind of causal map, though the encyclopedia does not yet list it as one.

Relationships to Other Abstractions¶

Current abstraction Causal Map Domain-specific

Parents (1) — more general patterns this builds on

Causal Map is a kind of Network Prime

An attributed causal map is a directed network of factors and asserted influence links.

Hierarchy path (1) — routes to 1 parentless root

Causal Map → Network → Reservoir-Flux Network → Conservation Laws → Invariance

Neighborhood in Abstraction Space¶

Causal Map sits in a moderately populated region (49^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Rhetorical Fallacies & Persuasion Tactics (24 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Concept map may include noncausal relations. Causal loop diagram uses feedback polarity conventions. Causal-inference DAG is typically acyclic and tied to structural assumptions for identification. Theory of change may use a causal map but also includes intervention goals, assumptions and evaluation commitments.^[1]

References¶

[1] Robert Axelrod, ed., Structure of Decision: The Cognitive Maps of Political Elites, original text preview, especially chapter 1 on causal assertions and coding. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n

[2] Colin Eden, “Strategic thinking with computers”, original article record and abstract, Long Range Planning 23 (1990). registry ↩a ↩b ↩c ↩d