A cause is a factor that helps produce, enable, prevent, accelerate or redirect an outcome.
The word cause is often used too loosely. A spark may trigger a fire, dry vegetation may enable it, wind may accelerate it, poor maintenance may contribute to it, and an absent firebreak may explain why it spreads. These causal roles are related but not identical.
Quick answer: how should causes be categorised?
- Role: trigger, enabling condition, contributing cause, necessary cause, sufficient cause?
- Proximity: immediate, intermediate or distal?
- Mechanism: how does the factor produce the effect?
- Timing: before, during or accumulated over time?
- Control: controllable, influenceable or external?
- Evidence: observational, experimental, historical or model-based?
- Confidence: established, probable, plausible or disputed?
- Interaction: independent, additive, synergistic or conditional?
- Scope: one case, a class of cases or a system-wide pattern?
This page focuses on classifying causal roles. For broader relationship structure, see How to Categorise Relationships; for explanatory accounts, see How to Categorise Explanations.
Cause is not the same as correlation
Two variables can move together because one causes the other, because both share another cause, because the apparent relationship is accidental, or because measurement and selection created the pattern. Correlation can suggest a causal question but cannot by itself settle it.
Triggering causes start transitions
A trigger is the event or threshold crossing that immediately starts a change. It may be dramatic and visible, but it can still be a small part of the deeper causal structure.
Enabling conditions make outcomes possible
Fuel enables combustion; network access enables a remote transaction; prerequisite knowledge enables advanced learning. Enablers can be essential without directly initiating the outcome.
Contributing causes increase likelihood or magnitude
Complex outcomes rarely have one sufficient cause. Several factors may each increase probability, duration or severity without being decisive alone.
Necessary causes must be present
If an outcome cannot occur without a particular factor under the stated model, that factor is necessary. Necessary does not mean sufficient.
Sufficient causes can produce the outcome under stated conditions
Sufficiency always belongs to a defined context. What is sufficient in one environment may fail in another because background conditions differ.
Proximate causes sit near the outcome
They explain the immediate mechanism: a fuse blows, a student misreads a command word, a server runs out of memory.
Distal causes shape the environment
Training gaps, institutional incentives, ageing infrastructure or historical policy can create conditions in which proximate causes become more likely.
Root cause should not become a ritual label
Many systems have multiple interacting causes. Calling one factor the root cause can erase enabling conditions, latent weaknesses and governance failures. It is often more accurate to record a causal structure than one privileged root.
Mechanism distinguishes causal stories from causal evidence
A plausible mechanism explains how one factor could produce another. But plausibility alone is not proof; the mechanism must still be supported by evidence relevant to the case.
Temporal order is necessary but not enough
A cause must precede or coincide with the effect in the relevant sense, but simply occurring earlier does not establish causation.
Accumulating causes act over time
Fatigue, corrosion, debt, skill decay and repeated small errors can accumulate until a threshold is crossed.
Protective causes reduce probability
Redundancy, training, vaccination, firebreaks or verification checks can causally reduce adverse outcomes. Causal classification should therefore include preventive and inhibitory relationships, not only producing ones.
Interactions matter
Two causes can be additive, redundant, synergistic or conditional. A factor may matter only when another factor is present.
Controllable causes differ from external causes
For action, it matters whether a cause can be removed directly, influenced indirectly, buffered against, or only monitored.
Case causes differ from population causes
A factor can explain why one case occurred without being the dominant driver across a whole population. Scope should always accompany causal claims.
Evidence should match the causal claim
Experiments, natural experiments, longitudinal data, mechanism evidence, comparative cases and historical reconstruction offer different kinds of causal support. Stronger claims require stronger designs and clearer assumptions.
Counterfactuals clarify causation
Ask what would likely have happened if the proposed cause had been absent or different. This helps distinguish genuine causal leverage from mere sequence.
Confounding creates false causal stories
A third factor can influence both supposed cause and outcome. Classification should preserve known confounders and whether they were controlled, measured or unresolved.
Mediation shows causal pathways
A cause can act through intermediate states. Preserving these steps improves explanation and intervention design.
Moderation changes causal strength
The same cause can have different effects under different ages, environments, system loads or institutional conditions.
Causal confidence should remain local
We may be highly confident that A contributed to B while uncertain about the size of the effect or exact mechanism. These uncertainties should not be collapsed.
Causal records need provenance
Record who asserts the relationship, what evidence supports it, which model was used, which alternatives remain and when the assessment was made.
AI can generate causal language too easily
Language models can produce fluent explanations that overstate cause. Treat extracted or generated causal relations as claims requiring evidence rather than as facts.
A practical cause record
- cause ID;
- affected outcome;
- causal role;
- proximity;
- mechanism;
- timing;
- necessary or sufficient status;
- interacting causes;
- controllability;
- supporting evidence;
- challenging evidence;
- confounders;
- confidence;
- source and version.
The deeper idea
Good causal classification prevents the nearest visible event from absorbing all explanatory attention. It makes the producing structure inspectable.
To categorise a cause well is to know what causal role it plays, how close it sits to the outcome, what mechanism connects it, what evidence supports it, and what intervention becomes possible if the relation is real.
Final answer
Categorise causes by causal role, proximity, necessity, sufficiency, mechanism, timing, interaction, controllability, evidence, confidence and scope. Keep correlation separate from causation, preserve multiple contributing causes in complex systems, and tie causal claims to the evidence and counterfactual reasoning that justify them.