There are competing candidates on what causation is, viz., regularity, counterfactual dependence, probabilistic dependence, process, etc. The literature on the metaphysics of causation typically operates under the presumption that there is a single, deep metaphysical fact about what causation is. Casual pluralism denies this presumption.
Analysis of causation often proceeds by thinking up intuitive examples involving causation and testing it against a causal theory. When this analysis fails either there is either (1) a rejection of the intuitions of the counterexample, or (2) acceptance of the counterexample and the addition of extra clauses in the causal theory to accommodate the problem. Arguably, the result is that an intuitive theory becomes modified to be less intuitive, ad hoc, and convoluted. This process can be illustrated by the massive literature on the counterfactual theory of causation including pre-emption, early pre-emption, late pre-emption, trumped pre-emption, double prevention, etc.
Persistent failure of these analyses raises the question as to whether there is a single, deep metaphysical fact about causation. The alternative is causal pluralism, and, appropriately, there is a plurality within causal pluralism.
Ned Hall puts forth a two-concept view of causation. On this view there are two concepts of causation.
- Dependence: the effect depends on the cause nomologically, or counterfactually, or probabilistically.
- Production: the cause produces the effect by a mechanism. This can be described by a transfer of a property, or a physical quantity.
You can have causal dependence without causal production and vice versa. For example, suppose air traffic controller sees two planes on a collision course, but you prevent him for sending the warning signal. You would be the cause of the collision on a dependence account but not on a production account. In another example, suppose two assassins fire at a target; you would produce the death but the death does not counterfactually depend on you since it is overdetermined.
Questions about general principles of causation depend on which of the two views you hold. Omissions can be involved in causal relations on the dependence view (e.g., lack of oxygen causing a drowning) but not on the production view. Transitivity, locality, and intrinsicness apply to the production view, but not necessarily on the dependence view.
On a cluster-concept view, there are multiple distinct criteria, which serve to identify causation. For example (taken from Psillos):
- there is a law (deterministic or statistical) that links c and e
- if c hadn’t happened, e wouldn’t have happened
- prob(e/c)>prob(e) in (all) relevant background contexts
- some causal process (mechanism) connects c and e
- something gets transferred from c to e.
If any particular criteria is satisfied, a relation could be identified as a cause. Causation would be an umbrella term. But, being pluralist, there is no deep essence, or single necessary criteria for causation here.
An Anscombe and Wittgenstein inspired view
On this view causal talk is dispensable for the most part. Causal language can be replaced by verbs like: scrape, push, wet, carry, eat, burn. Causal language adds no new content. For example, compare “Jones was shot to death” and “the cause of Jones death was a shooting.” The latter causal sentence adds no new content.
Compare the similarity of the verb ‘cause’ with the verbs ‘enable’ and ‘prevent.’ We don’t expect to discover any deep metaphysical fact about ‘enable’ and ‘prevent’, so why should there be one for ‘cause’?
Causal language can still be useful for generalizations or as a placeholder for when we currently lack a verb. For example, before we knew about electricity, we would say, “The death of Jones was caused by lightning.” Now, we could say, “Jones was electrocuted.” Whether there are sentences where a verb besides ‘cause’ wouldn’t work is unclear to me.
As a consequence of this view, you can directly observe causes, because you can directly see the wind bending the tree branch (or in causal language, you can directly see the wind cause the tree branch to bend.)
A worry is that there must be some deep metaphysical fact among the uses of all these verbs. One reply is to say that while there are things in common among these verbs that distinguish them from correlation, there could be multiple principles behind these uses, like counterfactual dependence or transfer of conserved quantity etc.
Essentially Contested Concept
The essentially contested concept view is held by Peter Godfrey-Smith. The idea is that the concept of ‘cause’ is not something that can be analyzed. The domain of applicability of the concept is susceptible to change, which means there needn’t be a fixed analysis (e.g., the concept ‘art’).
As an example of how the concept of ‘cause’ may change, imagine a situation where Jones didn’t care to save a drowning baby. We might say that Jones was responsible the drowning, even though there wasn’t causal interaction. In other words, we separate responsibility from the concept of ‘cause’. But, on the other hand, it might be more satisfying to say that he still did cause the drowning. Our use of ’cause’ may vary according to our culture.
Godfrey-Smith, Peter. Causal Pluralism.
Psillos, Stathis. Causal Pluralism.