Philosophy of Causation

What it asks

One billiard ball strikes another and the second moves. Everyone knows what happened. Say what makes it true.

Hume looked and found only three things: the first event, the second event, and the fact that events like the first have always been followed by events like the second. He could not find the causing. That negative result is where the branch begins, and every theory since is an attempt to say what causation is given that you cannot observe it directly.

The recurring problems

Distinguishing cause from correlation. Night follows day with perfect regularity and does not cause it. Any regularity theory needs to exclude cases like this without excluding real causes, and the fixes get baroque quickly.

Pre-emption. Two assassins; the first shoots and the second would have if the first had not. The victim would have died either way, so the counterfactual test says the first shot was not the cause - which is wrong. Repairing this has occupied the field for fifty years.

Absences. The gardener's failure to water killed the plant. If absences can cause, then so did everyone else who did not water it, and the causes of anything become innumerable.

Whether it is one relation. The cause of a war, of a fracture, of an economic collapse and of a chemical reaction may have nothing structurally in common beyond the word.

How it developed

Aristotle's four causes are not answers to this question. They are four kinds of explanation - what it is made of, what shape it has, what brought it about, what it is for - and only the third is causation in the modern sense.

Hume's analysis in the eighteenth century created the modern problem by removing necessity from the picture and leaving regular succession in its place. Kant's response was that causation is a category the mind brings to experience rather than something read off it.

The twentieth century produced three serious analyses in succession. The regularity theory refined Hume. David Lewis's counterfactual theory in 1973 replaced it with a much sharper test: the effect would not have occurred without the cause. Probabilistic theories dropped the requirement of certainty, which was necessary once medicine and social science became the paradigm cases rather than billiard balls.

The most recent turn is interventionist: a cause is something such that changing it would change the effect. That treats causation as bound up with what can be manipulated, and it has been unusually productive because it maps onto how experiments actually work.

The positions within it

Regularity theories, counterfactual theories, probabilistic theories, process theories and interventionism are the standing analyses. Causal eliminativism holds that fundamental physics contains no causation and that the notion belongs to the middle-sized world only.

All are treated in the Schools section.

Where it stands now

The interventionist framework has become dominant, largely because it works: causal modelling built on it is now standard equipment in epidemiology, economics and machine learning.

Whether it is an analysis of causation or a useful method that presupposes it is exactly the live philosophical question, and it is a good example of a branch where the practical success and the conceptual problem have come apart.