I start by trying to define two notions, cluster kinds and causality.
A cluster kind is when we recognize that there are exceptions to a generality, but we still try to use it in how it mostly works. Examples are saying that fruit is sweet (in spite of lemons), or when we say that birds fly (in spite of penguins and ostriches). There are certain limitations on what can be done with a cluster kind since basically we are removing the word “all” from the general statement. That means that we cannot use it syllogistically, as in saying, “All birds fly. Oscar is a bird. Therefore, Oscar flies.” But within certain recognized constraints, cluster kinds can still be useful. For instance, we might be able to describe a single situation by using multiple cluster kinds, so that reality lies in the overlap. (Oscar might also have feathers, wings, and a beak…, but not necessarily all).
Another way of putting it is that a cluster kind describes a collection of features which are mostly applicable to a given situation, but not necessarily all of them have to apply in every case. But there is still much that can be done with that, if we are willing to accept a degree of uncertainty to our conclusions. Cluster kinds are a way of trying to preserve “natural kinds”—even when the latter fail to break evenly into categories or fail to “carve nature neatly at the joints”—and so they are a way of trying to maintain realism.
A general discussion of cluster kinds can be found from Slater (see references, below).
Causality, on the other hand, has a long history of being hard to define, over centuries of trying, which is why it is worth considering that perhaps the problem is that causality is really a cluster kind. That would mean that causality is a collection of notions, not all of which have to apply in every case, and that that is what is making it difficult to give it a single definition. Ben-Menahem has suggested as much in her book Causality in Science (2018).
I am not here going to make an encyclopedic list of all the historical attempts to define causality, but some familiar ones include using counterfactuals; tests of manipulation; and Hume’s notion that causality must be “regularly” observed in order to be recognized by inductive inference.
This problem of defining causality has been stated well by Norton when he argues that causality is such a “plastic” notion—it can be bent into a lot of meanings—that it is “physically inept.” Or as I would put his argument, this plasticity make it useless as a yardstick by which to consistently judge other phenomena. It is like having a yardstick where we can adjust the spacing between the markings to give us whatever answer we want, rather than having a reliable measurement. So Norton speaks of causality as being on the order of “a folk science.”
Yet in so doing, Norton calls causality “a loose and varying collection of causal notions.” That is an undesirable quality in a standardized yardstick, but it sure does sound like a cluster kind.
So what about that? Does being a cluster kind rescue causality from being a mere folk science? Or does being a cluster kind diminish much of the power that causality is said to have, as, for instance, it being a bringer-about of deterministic change?
Slater includes a discussion of how Boyd started a cluster view of kinds specifically regarding causal mechanisms. Slater actually argues that causality—or appeals to how things are used—is not really needed in order to recognize cluster kinds. But now the issue has become: Is causality itself a cluster kind?
I will give my own answers in my next post.
Ben-Menahem, Y. 2018. Causation in Science. Princeton, NJ: Princeton.
Norton, J. 2003. “Causation as Folk Science.” Philosopher’s Imprint 3 (4): 1-22.
Slater, M. 2014. “Natural Kindness.” The British Journal for the Philosophy of Science 66 (2): 375-411.