Causality

Causality is the relation between causes and effects — the principle that events in the world do not occur randomly or for no reason, but are produced by prior events through law-governed processes. It is one of the foundational concepts of science, philosophy, and everyday cognition. Every explanation, every prediction, every intervention in the world presupposes causality: when we explain why something happened, we cite its causes; when we predict what will happen, we apply causal laws; when we try to change outcomes, we manipulate causes.

The concept is also one of the most contested in philosophy. We never observe causality directly — we observe sequences of events, correlations, and regularities. The inference from these observations to causal conclusions is philosophically contested, practically indispensable, and scientifically central. No theory of causality commands universal assent, yet every scientific practice implicitly uses one.

The Humean Challenge

David Hume's analysis of causality (1748) is the unavoidable starting point. Hume observed that when we claim A causes B, we mean more than that A and B are merely correlated — more than that B regularly follows A. We mean there is a necessary connection: A compels B to occur. But Hume argued that no such necessary connection is ever observed. We observe A, we observe B, we observe that B follows A reliably in our experience. But the necessity — the compulsion that makes us say A *must* produce B — is never directly experienced. It is something we project onto regularities in nature, not something we read off from them.

This is the Humean regularity theory of causation: causality just is constant conjunction — A causes B means nothing more than that events of type A are regularly followed by events of type B in our experience. The necessity we feel is psychological, not metaphysical: we become habituated to the sequence and form the expectation that B will follow A.

The consequence is radical: we have no rational justification for believing the future will resemble the past (the problem of induction), and no metaphysical grounding for the causal necessity we attribute to natural laws. Hume did not conclude from this that causality does not exist — he concluded that it is a fundamental feature of human psychology, not of mind-independent reality.

Counterfactual and Interventionist Accounts

The most influential modern account, developed by David Lewis (1973) and refined by many subsequent philosophers, analyzes causation in terms of counterfactuals: A causes B if and only if, had A not occurred, B would not have occurred. This captures the intuition that causes are difference-makers: if removing the cause would have prevented the effect, the cause is genuine.

Counterfactual theories have two advantages: they align with how we actually reason about causality (we ask "would the accident have happened if the driver hadn't been drunk?"), and they explain asymmetry (causes precede effects, and the counterfactual runs forward in time). They face the problem of overdetermination: if two independent causes each would have been sufficient for the effect, the counterfactual test fails for each individually — neither is necessary — yet both intuitively caused the effect.

Judea Pearl's interventionist theory (2000) connects causality to manipulation and experiment. A causes B if intervening on A (holding everything else equal) changes B. This is operationalized through causal graphs: directed acyclic graphs (DAGs) in which nodes represent variables and directed edges represent causal relationships. Pearl's do-calculus provides a formal language for distinguishing the question "what is the correlation between A and B in the observed data?" from "what would happen to B if we intervene to set A?" — the distinction between observation and experiment that is the methodological foundation of causal inference in statistics and medicine.

The interventionist account has a clean connection to scientific practice: randomized controlled trials are precisely the gold standard because they implement the intervention operator — they set the value of the treatment variable while holding everything else random, blocking confounds. Observational data cannot, in general, support causal claims without additional assumptions, because correlation without intervention always admits confounding.

Causality and Physical Theory

Newtonian physics seemed to vindicate a robust metaphysical causality: the universe is a deterministic system of particles under forces, and every state is caused by the prior state through Newton's laws. Causality was absolute: if you knew the initial conditions and the laws, you could predict every future state, and trace every future state to its prior causes.

Quantum mechanics disrupted this picture. The collapse of the wave function upon measurement appears genuinely random — not determined by prior causes in any recoverable sense. The decay of a radioactive nucleus at a particular moment has no cause in the classical sense: given identical initial conditions, different decay times can occur. This raised the possibility that causality in the classical sense fails at the quantum level.

The response has been contested. Hidden variable theories (Bohm) attempted to restore causality by positing additional variables beyond the quantum state. Bell's theorem ruled out local hidden variable theories experimentally: the correlations in quantum entanglement cannot be produced by any local causal mechanism. What remains is a deeply non-classical causal structure in which remote measurements can be correlated in ways that classical causality cannot explain — though the correlations cannot be used to transmit information (no signaling theorem), preserving causality at the macroscopic level.

Special relativity provides a structural constraint on causality: the light cone. Events can only causally influence events inside their future light cone; causal influence cannot propagate faster than light. This is the sharpest physical version of causal direction.

The Causal Structure of Science and Culture

The essentialist's claim: causality is not merely a useful concept — it is the concept that makes science, explanation, and rational intervention possible. Every scientific explanation is a causal explanation, explicitly or implicitly. Every policy is a causal intervention. Every narrative — in history, in literature, in law — is organized around causal structure. Hume was right that necessary connection is not directly observed. He was wrong to conclude that causality is merely psychological. The success of causal reasoning in producing reliable predictions and effective interventions across every domain of human inquiry is strong evidence that we are tracking something real, even if the metaphysical nature of that real thing remains contested.

The hypothesis that the universe is causally structured — that events are connected by law-governed relations of production, not merely associated by habit — is the most successful empirical hypothesis in the history of science. It has produced technologies, medicines, institutions, and explanations that work. A theory of the world that eliminated causality in favor of mere correlation would make prediction possible but intervention unintelligible. We would know that smoking is correlated with cancer but could not conclude that stopping smoking would reduce cancer rates. The power of causal thinking is precisely that it supports not just prediction but action. Any account of science or culture that treats causality as dispensable has not thought through what it is dispensing with.