PHIL2109 Contemporary Metaphysics
Week 8: Causation

Causation

  1. There is such a thing as causation. The dropping of the bomb on Hiroshima, for example, caused the death of many people.
  2. There might be more than one thing that we mean by the word 'causation', in which case we might need to first disambiguate 'causation' before we start talking past each other. But we shall assume otherwise.
  3. We are interested in learning more about the nature of causation. Some might say that we are interested in learning more about our concept of causation. Perhaps we are, but only insofar as it tells us about the nature of causation itself.
  4. How can learning more about our concept of causation tell us about the nature of causation? One way is if the concept is complex, having other concepts as constituents.

    Example. Here are some things that we know about our concept BACHELOR: it is complex; it has, as constituents, our concepts NOT, AND, MARRIED, ADULT, and MALE; in fact, BACHELOR = (((NOT MARRIED) AND ADULT) AND MALE). From this we can conclude that all bachelors are males. In fact, we can conclude something stronger: necessarily, all bachelors are males. These last two things are not facts about our concept BACHELOR - they are facts about bachelors. But we can come to know them by learning more about our concept BACHELOR. They are conceptual truths: not truths about concepts, but truths guaranteed by concepts. We can arrive at them by doing conceptual analysis.

  5. Perhaps we can arrive at truths about causation in the same way. Perhaps this is the only way, given that we metaphysicians are confined to our armchairs. If it turns out that our concept of causation is simple, so that there are no conceptual truths to be found, then we may not discover anything at all about causation. At least not as philosophers.

Some questions

  1. Here are some things that we would like to know:

    We will consider just some of these questions.

The relata of causation

  1. Assuming that causation is a 2-place relation, what are its relata?
  2. Suppose that John explodes a bomb and thereby causes a bridge to collapse. He are some things that it seems true to say:
    1. John caused the bridge to collapse.
    2. John caused there to be a collapsed bridge.
    3. John caused the collapse of the bridge.
    4. John caused a mess.
    5. The exploding of the bomb caused the collapse of the bridge.
    6. The fact that a bomb exploded caused the collapse of the bridge.
    7. The bridge collapsed because the bomb exploded. (? This might not be an expression of a causal fact. It might be an expression of a causal explanation.)
  3. So we at least talk as if the following kinds of things can stand in the causation relation:

    This does not mean that these things actually can stand in the causation relation. But it does suggest that there is nothing in our concept of causation that rules it out. So we should probably take it that they can, at least until we have reason to think otherwise. (In the next section we will see one such reason.)

  4. An interesting question is whether or not one of these things is the primary kind of relatum.
  5. In any case, we shall focus on event causation: causation that obtains between events.

Fact causation

  1. Is there ever any fact causation? That is, is it ever the case that one fact causes another fact? (Set aside the question of whether this kind of causation is primary or secondary.)

    Note: this is different from asking whether there is such a thing as fact causation.

  2. Here are some possible examples:
  3. But we have the so-called slingshot argument:
    1. Suppose that the fact that I ate cake caused the fact that I was happy.
    2. So the fact that Sydney is the thing such that it is Sydney and I ate cake caused the fact that I was happy.
    3. So the fact that Sydney is the thing such that it is Sydney and there are kangaroos in Australia caused the fact that I was happy.
    4. So the fact that there are kangaroos in Australia caused the fact that I was happy.
    5. So, by similar reasoning, the fact that there are kangaroos in Australia caused the fact that Topeka is the capital of Kansas.

    So if any statement of fact causation is true, then every statement of fact causation is true. So no statement of fact causation is true.

  4. Two principles used in the argument:
    1. If the fact f1 caused the fact f2, and if f1* is logically equivalent to f1, then the fact f1* caused the fact f2. Similarly, if the fact f1 caused the fact f2, and if f2* is logically equivalent to f2, then the fact f1 caused the fact f2*.
    2. If the fact that a has p caused the fact f2, and if a = b, then the fact that b has p caused the fact f2. Similarly, if the fact f1 caused the fact that a has p, and if a = b, then the fact f1 caused the fact that b has p.
  5. Lowe talks about the idea that causation is an extensional relation. What does this mean?

How many causes?

  1. Here is a mess that needs sorting out:

    We sometimes talk about one event being a cause of another event, which suggests that events can have multiple causes. But we sometimes talk about an event being the cause of another event, which seems to presuppose that there is just one cause. And when we talk about one event causing another event, do we mean that it is a cause or the cause of that event?

  2. Note that if event causation is transitive then at least some events have more than one cause. Suppose that e1 caused e2 and that e2 caused e3. Then, by transitivity, e1 caused e3. So both e1 and e2 caused e3.
  3. Here is one possibility: events can have more than one cause, but sometimes when there are multiple causes we focus on just one of them and refer to it as 'the' cause. This is a general and much discussed issue in the philosophy of language. Compare: 'This is John's book'.
  4. Here is another possibility. There are sufficient causes (or: complete causes) and partial causes (or: contributory causes). c is a sufficient cause of e iff the occurrence of c causally necessitates the occurrence of e. These sufficient causes might be composed of partial causes.

    Suppose I light a match and there is an explosion. The lightning of the match is not by itself a sufficient cause of e or a complete cause of e - it was a contributory or partial cause of e. Other causal factors: the presence of flammable gas, the presence of oxygen. (These are not events, but something like conditions.)

  5. Events that occur at least partly by chance thus have no sufficient cause. This raises the issue of probabalistic causation. We will not discuss this (see Lowe).
  6. We could have two events that count as a cause of e, each of which is not a sufficient cause of e. This would not be a case of causal overdetermination. They are each partial causes of e.
  7. We could have two events that count as a cause of e, each of which is a sufficient cause of e. This would be a case of causal overdetermination.

The Humean analysis of causation

  1. Hume proposed the following constant conjunction analysis of event causation:

    Event c was a cause of event e iff (a) c preceded e, and (b) there are event types t1 and t2 such that: c is of type t1, e is of type t2, and every event of type t1 is followed by an event of type t2.

  2. Note that the analysans does not seem to invoke the concept of causation. (analysandum iff analysans)
  3. Problem. We need to restrict the types that are allowed, or else we get far too much causation (the analysis overgenerates). Take any two events at all, e1 and e2, such that e1 precedes e2. let t1 be the type of being identical to e1, and t2 be the type of being identical to e2. Then the analysis declares e1 to be a cause of e2.
  4. Problem. Suppose c was a cause of e. Suppose e is an event of type t2. Then c is an event of type t1: being a cause of an event of type t2. If causes always precede their effects then every event of type t1 is followed by an event of type t2. So the 'only if' part is trivially satisfied. Unless we add that the types are non-causal types.
  5. Problem. Suppose it just happens that every event of type t1 is followed by an event of type t2. t1: my sister having a baby; t2: some other woman having a baby.
  6. Related problem. The analysandum implies the counterfactual that if c had not ccurred then e would not have occurred, but the analysans does not.
  7. Problem. Why should other events matter to whether or not c is a cause of e?

The counterfactual analysis of causation

  1. Here is an analysis, call this the simple analysis:

    (SCA) Event c was a cause of event e iff (a) c and e occurred, and (b) if c had not occurred then e would not have occurred.

  2. Problem: Napoleon's birth was not a cause of Napoleon's death, but if Napoleon's birth had not occurred then Napoleon's death would not have occurred.
  3. Problem: if I had not written 'vodka' then I would not have written 'v'. If I had not written 'v' then I would not have written 'vodka'. So each is a cause of the other. But causation is asymmetric. Worse, neither is a cause of the other. Causation can only obtain between wholly distinct events - no parts in common.
  4. In response to this problem, we might try:

    (SCA+) Event c was a cause of event e iff (a) c and e are wholly distinct events (i.e. they have no event as a common part), (b) c and e occurred, and (b) if c had not occurred then e would not have occurred.

  5. Note: it is not part of the concept of causation that causes must precede their effects, so this cannot be part of a conceptual analysis. Could have backwards causation, and simultaneous causation.
  6. A problem. Suppose that pressing a button causes a bomb to explode. Then SCA+ declares the exploding of the bomb to be a cause of the pressing of the button (if the exploding had not occurred then the pressing would not have occurred).

    No good to stipulate that c occured earlier than e, because, even if backward causation is not possible, that is not something we know conceptually. Also: Suppose that pressing a button causes a warning light to go on and then a bomb to explode. SCA+ declares that the flashing of the light caused the exploding of the bomb. And this time it occurred before it.

    No good to rule out backtracking counterfactuals. But this again would rule out backward causation.

    Lewis: the counterfactual in question is just no false (no true reading).

  7. Also, we get problems of undergeneration from the following kinds of cases:
  8. What if we accept these as counterexamples to SCA+? Try this instead:

    (CCA) Event c was a cause of event e iff (a) c and e are wholly distinct events, (b) c and e occurred, and (c) a chain of counterfactually dependent events linked c to e.

  9. Note that counterfactual dependence is not transitive, so we might have made some progress here. But this can be questioned - Lowe in fact does so.
  10. Note that the new analysis declares causation to be transitive, too, which the old one does not.
  11. How is this supposed to overcome the problem of pre-emption?
  12. But we get a problem with cases of late pre-emption. These are cases in which the only effect of c that prevents the completion of a chain of counterfactually dependent events linking d to e is e itself. So CCA declares, wrongly, that c is not a cause of e.