Where does cause collide?

Aristotle held in the Physics and the Metaphysics (ca. 350 B.C.E.) that to know a thing is to grasp its aitia, the four answers to the question of on account of what it is: the material cause, what it is made from; the formal cause, the form or definition that makes it the kind of thing it is; the efficient cause, what brings it about; and the final cause, the end or purpose for which it exists. His word aitia is closer to "explanatory factor" or "the because" than to the modern idea of one event producing another. Hocutt argued in his 1974 Aristotle's Four Becauses that the doctrine reads better as a theory of explanation than of causation, since only the efficient cause resembles what later writers call a cause. A cause in this meaning is one of four kinds of answer to a why-question about a thing, not a set of facts that founds a lawsuit and not a measured effect of a treatment in a population.[1][2]

Unit

One of the four explanatory factors of a thing: material, formal, efficient, or final

Recovery means

Not applicable

Measured by

Identified by analysis of what a thing is made of, what it is, what produced it, and what it is for; no instrument

The set of facts that, taken together, entitle a person to a judicial remedy against another. The principle is old: in Ashby v. White (1703) Chief Justice Holt held that a man wrongly barred from voting had an action even without any money lost, on the ground that where the law gives a right it gives the means to vindicate it. Clark set out the modern construct in his 1924 The Code Cause of Action, defining it under code pleading as the group of operative facts giving rise to one enforceable claim, so that a single set of facts is one cause of action however many legal theories or remedies it might support. A cause of action in this meaning is the ground of a lawsuit, counted as one claim, and it is distinct from the separate question of causal link a court asks inside a claim, from the four explanatory factors of the philosophers, and from a treatment effect estimated in data.[3][4]

Unit

One enforceable claim, defined by the group of operative facts that give a right to sue

Recovery means

Not applicable, the aim is damages or other relief once the claim is established

Measured by

Tested on the pleadings: whether the alleged facts state each element of a recognised claim, so as to survive a motion to dismiss

The effect of a cause defined as a contrast between potential outcomes: what a unit's outcome would be under one condition against what it would be under another. Neyman introduced the potential-outcomes notation for randomised agricultural trials in a 1923 essay, translated into English in 1990, writing the yield a plot would give under each treatment and defining the effect as their difference. Rubin generalised the framework to observational as well as experimental studies in his 1974 Estimating causal effects of treatments in randomized and nonrandomized studies, and Pearl's 2009 Causality added the do-operator and graphical rules for when such an effect can be identified from data that were only observed. A cause in this meaning is a manipulable condition whose average effect is estimated across a population, not one of the four explanatory factors of a thing and not the group of facts that founds a lawsuit.[5][6][7]

Unit

A treatment or condition applied to units, whose effect is the difference in their potential outcomes

Recovery means

Not applicable, the target is identification: whether the average effect can be written from the observed distribution

Measured by

Estimated as an average treatment effect from randomised or covariate-adjusted comparisons, with identifiability judged by graphical criteria