Causation

Causation is the relationship through which something produces, influences, or makes a difference to something else. Causes and effects may be understood as events, facts, conditions, or processes, depending on the theory adopted. Within philosophy, causation raises questions about what connects occurrences, how causal relationships can be known, and whether they can be explained through more basic concepts. Its investigation therefore crosses metaphysics, epistemology, and the philosophy of science. No single analysis commands universal agreement. (plato.stanford.edu)

Causation and association

Causation differs from mere succession or statistical association. One event’s following another does not establish that the earlier event produced it. Likewise, correlation may arise because two variables share a common cause, because influence runs in the opposite direction from that supposed, or because of the way observations were selected. A confounder, for example, can influence both a proposed cause and its apparent effect, creating an association that does not represent the proposed causal relationship. (ftp.cs.ucla.edu)

Philosophers also distinguish singular causation—what caused a particular occurrence—from general causal relationships between kinds of events or variables. A general claim need not imply that the proposed cause invariably produces its effect. Probabilistic theories use probability to describe causal influence without requiring deterministic sufficiency: a cause may change an effect’s likelihood rather than guarantee its occurrence. (plato.stanford.edu)

Historical approaches

In ancient Greece, Aristotle developed the doctrine of the four causes. Material causes concern what something is made from; formal causes concern its form or organization; efficient causes identify sources of change; and final causes concern ends or purposes. This framework addresses several kinds of explanatory question, rather than only the event-producing relationship emphasized in many modern discussions. For Aristotle, understanding something scientifically required grasping its appropriate causes. (plato.stanford.edu)

David Hume examined the basis of causal expectation. Observing one event followed by another, he argued, does not reveal a necessary connection between them. Repeated conjunction instead generates a habit of expecting the familiar effect. His discussion connects causation with inductive reasoning: past regularities alone do not demonstratively establish that the future must resemble the past. Interpretations differ over the broader metaphysical implications of his position. (plato.stanford.edu)

Immanuel Kant treated causality as an a priori category of understanding, rather than a concept simply abstracted from repeated observations. In his account, experiencing an objective succession of events requires their connection according to a rule. The disagreement concerns not merely particular causal judgments but the conditions under which experience can represent an ordered world. (plato.stanford.edu)

Major philosophical theories

Regularity theories connect causation with patterns governed by laws or appropriate background conditions. Their difficulty is distinguishing genuinely causal regularities from accidental correlations and relationships generated by common causes. Mere repeated co-occurrence is therefore insufficient without further qualifications. (plato.stanford.edu)

Counterfactual theories analyze causal relationships through alternatives to what actually happened. A simple counterfactual test asks whether the effect would have occurred if the proposed cause had not occurred. David Lewis developed an influential account using comparisons between possible worlds and chains of causal dependence. Such theories must specify which alternative circumstances remain fixed and which may change. (plato.stanford.edu)

Probabilistic theories characterize causes as changing the probabilities of effects under suitable conditions. A simple probability-raising test is inadequate because noncausal associations can also raise conditional probabilities. More elaborate accounts therefore consider background circumstances, common causes, and distinctions between observational and causal dependence. (plato.stanford.edu)

Interventionist theories connect causation with what would happen if a variable were changed through a suitable intervention. The intervention must isolate the proposed causal influence rather than alter the outcome through an independent route. Such accounts need not require actual human manipulation: hypothetical interventions can illuminate relationships involving events people cannot control. (plato.stanford.edu)

Process theories emphasize the physical connection linking cause and effect. Some accounts identify causal processes and interactions through conserved quantities such as energy or electric charge. They seek to distinguish genuine physical influence from merely correlated changes, although their treatment of absences and prevention remains contested. (plato.stanford.edu)

Formal causal inference

Causal inference develops methods for drawing causal conclusions from data together with explicit assumptions. A structural causal model represents how variables depend on other variables and background factors. In acyclic models, these relationships can be displayed using a directed acyclic graph. Arrows express assumed causal structure, not merely observed association. (ftp.cs.ucla.edu)

The expression $P(Y\mid do(X=x))$ denotes the outcome distribution when an intervention sets $X$ to $x$. It generally differs from $P(Y\mid X=x)$, which describes observations where $X$ already has that value. Randomized controlled trials can help separate causal influence from pre-existing differences, while observational inference requires justified assumptions. Identification asks whether those assumptions and the observed distribution uniquely determine the causal quantity sought. (ftp.cs.ucla.edu)

Difficult cases and conceptual limits

Overdetermination and preemption challenge simple counterfactual tests. In overdetermination, multiple actual causes may each suffice for an effect, so removing either alone leaves the effect unchanged. In preemption, one causal process produces the effect while a backup process would have produced it otherwise. These cases motivate refinements that distinguish actual causal contribution from unused alternatives. (plato.stanford.edu)

Omissions and prevention pose different problems. Whether failing to perform an action counts as a cause can depend on which alternatives are considered relevant. Prevention may explain an outcome’s absence without supplying a continuous physical process leading to that absence. These issues expose a persistent tension between understanding causation as productive connection and understanding it as difference-making. (plato.stanford.edu)