This paper is, for the most part, a tutorial on the concept of causality. It discusses several different kinds of causality, and attempts to clear up common misunderstandings associated with the different kinds. It also highlights shortcomings in the vocabulary that we use to discuss causality. For example, if A caused B and B caused C, we could say that A caused C. But, if A and B are independent, but together A and B caused C, we would still say that A caused C, even though the nature of the causality would be quite different.

The target audience for this paper is researchers and practitioners in data mining, because the purpose of data mining is to uncover relationships between data items that may be causal in nature. However, the paper is probably of more general interest, since causality is an important concept in all research, and is generally rather poorly understood.

This is an extremely important topic, and although the author does a good job of increasing the reader’s awareness of the complexity of causality, it is unfortunate that the author did not invest a little more time in improving the quality of the treatment. For example, it was the philosopher David Hume who was the first to question the concept of causality, and yet there is no mention of Hume anywhere in this paper. The author does say, “Philosophers, mathematicians, computer scientists, cognitive scientists, psychologists, and others have formally explored questions of causation beginning at least three thousand years ago with the ancient Greeks.” Unfortunately, none of the philosophers, mathematicians, or even ancient Greeks are mentioned, much less cited. Later in the paper, the author claims “there are at least four types of causality,” but then provides only two. Again, later in the paper, when the author is describing the asymmetries of time order, he states, “Effects do not come before effects.” No doubt he meant effects do not come before causes, but this is just sloppy, and detracts from an otherwise valuable paper.