What happened?
Candidate cause and outcome.
Causal claims require clear events, states or variables.
Side 94
Causation
A study of what it means for one thing to make a difference to another. Unlike Causal Inference, which asks how effects can be estimated from evidence, Causation asks what causal relations themselves are.
cause→dependence→mechanism→intervention→counterfactual
06causal lenses
05theory families
05hard cases
94Side
A cause is something whose presence matters to what happens.
That intuitive idea becomes difficult once multiple causes, prevention, omission and preemption enter the picture.
Would the outcome change if the cause changed?
Dependence.
Difference-making captures why causes matter rather than merely precede.
How is influence transmitted?
Mechanism or process.
Mechanistic accounts explain what links cause and effect.
Would changing the cause change the outcome?
Manipulability.
Intervention formalizes causal control while distinguishing it from observation.
Which background conditions matter?
Causes are rarely sufficient alone.
Causal status often depends on a surrounding structure of enabling conditions.
Counterfactual theories define causation through what would have happened otherwise.
A causes B when, under the relevant alternative where A does not occur, B would not occur—or would occur differently.
A and B occur.
The observed sequence alone does not establish the causal relation.
Imagine A absent or changed.
The comparison should preserve relevant background conditions.
B differs under the alternative.
This difference provides the core counterfactual relation.
Which possible world counts as closest?
Counterfactual truth depends on how alternatives are selected.
Dependence can fail despite causation.
A backup cause can make B occur even if the actual cause were absent.
Several causes may each be sufficient.
Simple but-for dependence can miss genuine causal contribution.
Mechanistic theories ask what carries the influence.
Causation becomes intelligible when entities and activities connect initial conditions to outcomes through an organized process.
Entities
What participates?
Molecules, cells, people, institutions or components can occupy mechanistic roles.
Activities
What do they do?
Binding, transmitting, signaling or deciding are process elements.
Organization
How are parts arranged?
The same parts can produce different outcomes under different organization.
Intervention
Does changing one component alter the process?
Manipulation helps test whether a proposed mechanism is causally relevant.
Levels
Which explanatory scale?
Mechanisms can be decomposed further without making higher-level explanation useless.
Intervention distinguishes causal from merely predictive relationships.
If changing X while breaking its usual causes changes Y, X occupies a causal role in the modeled system.
| Relationship | Observation | Intervention | Causal implication |
|---|---|---|---|
| Correlation | X and Y vary together | Unknown | Insufficient alone |
| Common cause | X predicts Y | Changing X may not change Y | Association can be spurious |
| Mediation | X relates to M and Y | Changing pathway alters Y | Mechanism becomes testable |
| Direct effect | X predicts Y after controls | Manipulating X changes Y | Supports causal role |
| Feedback | X and Y influence each other | Direction-specific intervention needed | Causal structure is cyclic |
Some theories treat causation as more than regularity or dependence.
Powers and dispositions describe entities as having capacities that can manifest under suitable conditions.
Fragility without breaking.
An object can possess a causal disposition even when it never manifests.
Capacity to bring about change.
Powers views treat causal potential as part of what entities are.
Regular structure constrains behavior.
Law-based theories connect causation to nomological patterns.
Causes are followed by effects.
Humean views resist adding hidden necessary connections beyond patterns.
Cause generates effect.
Production language captures something ordinary dependence accounts can feel too thin to express.
Causes can change chances.
A cause may increase the probability of an outcome without guaranteeing it.
Hard cases expose weaknesses in simple causal definitions.
A mature theory must handle omission, prevention, redundancy and causal chains without counting every correlation as causal.
Preemption
One cause produces the outcome before a backup cause can do so.
Overdetermination
Several independent causes are each sufficient for the same outcome.
Omission
Failure to act can appear causally relevant despite no positive event occurring.
Prevention
A cause can produce the non-occurrence of another event.
Probabilistic causation
A cause changes chances without determining the result.
Causationdifference-making and production
CounterfactualsDavid Lewis · dependence and possible worlds
Causal Modelsintervention and structural causation
Mechanismsentities, activities and organization