“Mildred, why did your cat die?”

“Because John didn’t feed it when I was out of town.”

Is this a philosophically valid account of causation? Not according to Helen Beebee!

In her paper Causing and Nothingness she claims that the absence of an event cannot be a cause1. This is a strong claim because, as in our example above, we refer to absences as causes in common language all the time. If we are to throw away such a major point of our common-sense understanding of causation, there had better be a good reason.

What is causation?

She upholds instead a “network model” of causation, in which the casual structure of the universe is a DAG, events are nodes and causal relations are directed edges from causes to effects. (She doesn’t insist that the graph be acyclic but I’m assuming that’s in there.) Now, critically, absences cannot be nodes in the graph because they are not events.

Events typically have a time and place where they happen. When exactly did John fail to feed the cat? Was it Tuesday at 7, or Wednesday at 8? Where was the event? Was it at Mildred’s house where the cat was, or was it at the bar where John went instead?

I don’t find this entirely persuasive, since there are events that have similar problems: “Billy grew up”. That happened, and yet…is there a time and place you can point to? There was a process of Billy growing up, just as there was a process of John not feeding the cat. I also don’t understand what is the granularity of the events in the graph: “John shot my cat” could be one event, but also you could break that down into a sequence of sub-events: John pulled the trigger -> the hammer struck the bullet -> etc. Both of these are valid, but is there a “true” structure? Do we need knowledge of each subatomic particle to create the “true” causal network of the world?

Are any absences causes? Which ones?

But back to Beebee’s account. She gives another problem with absences as causes: there are too many of them! Steve, Ed, and Tom Hanks all failed to feed her cat as well. Also, her cat died because it didn’t grow hands to open the cupboard doors and feed itself. Do we say that all these, too are causes? What distinguishes John’s failure such that our common sense calls that a cause but not the others?

We can imagine a definition where an absence A causes an event E if:

  1. Had A occurred, E would not have occurred, and
  2. The absence A is either abnormal or violates some norm (eg John promised Mildred he would feed the cat, or the automatic cat food dispenser usually works, or cats don’t usually grow hands).

This tracks with common sense, but Beebee doesn’t like it. Condition (2) is a matter of judgment and human-dependent norms, and it’s unfulfilling to say that they can affect the true causation in a universe made of uncaring atoms and physical law. And (2) is also a relative, not an absolute standard: Beebee wants to be able to discover a condition that separates events into causes and non-causes.

Causal statements and causal explanations

Okay, so we’re back to the original problem. Another solution is to define a thing called a “causal explanation” which somehow describes the true causal network, but less rigorously. A causal statement might be “A causes B”, which means that A and B are nodes in the graph and there is an edge from A to B. A causal explanation on the other hand might be something like “S, a category which includes A, caused B.” Or you can say “C did not cause B”, meaning that there is no event C upstream from B in the causal graph. So causal statements are a subset of causal explanations.

So then, when you say “the absence A caused event E”, what you are doing is actually giving a causal explanation by describing a “nearby world” with a causal graph where the event A actually occurred and was a cause of E.

Areas for further thinking

This is all a fully deterministic universe. How does it change if we think that there is randomness in the world as well as the model?

She draws a sharp line between this world and nearby ones, saying that “causal statements” are only true about this world. But this leaves open a theory of what “nearby worlds” actually are, and how they are connected to this one. Why is describing a nearby world relevant to us, who live in this one? What purpose does that serve? Is it possible to make a statement with a high level of rigor about these nearby worlds?

I learned about this paper on this episode of the excellent Journal Entries podcast which I’ve recommended previously here and here. The episode page has links to further reading as well: I want to explore these ideas more.

  1. Actually her goal in this paper is a bit different: “The central claim of this paper is that, if our aim is to do as much justice as possible to common sense intuitions about causation by absence, then it doesn’t much matter whether we uphold relationism or non-relationism.” But for our purposes, which is to explore the ideas rather than jump into an existing philosophical debate, this is close enough.