Thomson on Harm and Harming

At her paper at the Rocky Mountain ethics conference, Judith Jarvis Thomson discussed various accounts of the metaphysics of harm. Somewhat surprisingly, she accepted the following equivalence.

  • A harms B iff A causes B to suffer a harm.

Even more surprisingly, she defended this by saying it was a general claim about how causal verbs work. But this isn’t at all how causal words work. Compare this claim.

  • A breaks B’s window iff A causes B’s window to be broken.

Here’s a counterexample to that. A is a speaker at a philosophy conference. She makes an outrageous claim about the semantics of causal verb. This so upsets C that he storms out of the room, and in his anger punches the window of B’s car. The window breaks. Now it seems clear that A has caused B’s window to be broken, with of course some help from C, but A didn’t break B’s window.

So I was thinking that the biconditional about harming and causing harms would also be false. And I was thinking that cases of indirect causation, like this one, would be examples of when they were false. But when I wrote up the case, it became less clear.

So question: In the case just described, where C breaks B’s window, does A harm B? It’s clear that A does cause B to suffer a harm. And if pushed I would say that A didn’t harm B – that only C harmed B. But my intuitions are nowhere near as clear as I hoped. What do you think?

5 Replies to “Thomson on Harm and Harming”

  1. There’s a draft paper version of this talk in which Thomson explicitly acknowledges the point you’re making. She never endorses a biconditional like the one you have above, but instead only says that to harm is “roughly” to cause to be harmed. Later in the paper, she explains why she hedges:

    “But why all those “roughly”s? I postponed description of the complexity I mark by using them until we had the comparative as well as the noncomparative causal verbs before us, since what I use the “roughly”s to mark is common to all of the causal verbs. The complexity I refer to issues from what might be called considerations of agency.

    It would not in fact be true to say that for X to kill A is for X to cause A to be dead. In certain contexts, a person who says the words “B killed A” may properly be taken to assert a proposition such that it is enough for it to be true that B caused A to be dead – as, for example, where what isunder discussion is who is responsible for A’s death. (During the Vietnam War era, students chanted “LBJ, LBJ, how many kids did you kill today?” But I doubt that they thought LBJ had shot any kids that day.) In other contexts, a person who says the words “B killed A” may properly be taken to assert a proposition that is true only if B himself carried out the enterprise of killing A. (“I didn’t kill him, though I had him killed,” may very well be true.) Neither of these understandings of the words is the correct
    understanding of them: both understandings are correct, each in its place.”

  2. I wonder whether a generalization of this issue can shed some light on the special case of harming. Consider the following general principle:

    A Fs B iff A causes B to receive (or experience, or undergo, or suffer, or…) F (or F-ing).

    For what Fs, if any, is this principle worth considering? Indeed, for what Fs is the hedged version of this principle worth considering? Pushing, hating, creating, comprehending, reaching, befriending, observing, envying, etc… all run into counter-considerations of the same type as the one Brian brings up for breaking B’s window.

    What appears to matter to our reactions is not just the identity of F, but the identities of A and B, as well as their relationship. In the case of harming, I’d wager more people would agree that Brian’s case fits Thomson’s biconditional if we add that A is B’s parent, for example, or that A has a way to reliably predict C’s actions…

    But perhaps the general principle of which Thomson’s harm biconditional is an instance can be augmented, making room for precisely that kind of considerations. Consider, for instance:

    A Fs B iff A is the prominent part of the causal ancestry of B receiving (or…) F (or F-ing).

    Note that “prominent” needn’t be just another hedge word. It could well turn out that prominence conditions depend on, and co-vary with, the category or type to which the F at issue belongs. For example, the type relevant to moral evaluation could have different prominence conditions from, say, the type relevant to practical, “billiard ball” deliberation. Thomson’s biconditional, for instance, becomes:

    A harms B iff A is the prominent part of the causal ancestry of B suffering harm.

    Perhaps considerations of agency Dan mentions are the way to go, but even so the application of the above template to harm could, perhaps, better keep Thomson’s discussion on target than her original biconditional. The real difficulty would then be to specify and explain the differences between all the various types of prominence.

    Indeed, perhaps even the imprecise, commonplace sense of causing (if it differs from the precise sense of much of the work on the metaphysics of causation) itself fits the last template:

    A causes (in the commonplace sense) B iff A is the prominent part of the causal (in the metaphysically precise sense) ancestry of B undergoing causing (in the commonplace sense).

    The prominence of commonplace causation would, presumably, differ from the prominence of moral evaluation. (In fact, if combined with a distinction between two senses of causing, that could help explain why Thomson’s harm biconditional strikes us as problematic.)

  3. The position seems to entail that in any case in which you have to distribute a limited resource you harm someone. That can’t be right. So I have a limited cure for some disease and there is a just distribution of the cure D. If I choose D and someone suffers more than he would have otherwise, do I harm him? It’s hard to see how.

  4. Jerry Fodor has a bit about this in chapter 3 of his book “Hume Variations” where he wants to pose a question about putatively causal situations which he takes to be prior to the disagreement between rationalists and empiricists on the status of concepts that don’t self-evidently derive from experience. (Briefly, what fills the gap, e.g., between the impression of billiard ball A striking ball B and the thought that A moves B?)

    Fodor says that everyone admits that there are plenty of counterexamples to the principle that a proposition involving causing an intransitive – e.g., x causes (y move_I ) – implies a proposition involving the causal transitive – e.g., x moves_T y. The problem, as Brian Weatherson points out, is that nothing prevents some intermediary from actually doing “the dirty work” since x can cause y to move by getting z to move y. Fodor’s claim is that his friends and colleagues over twenty or thirty years have been unable to give a satisfactory defense of the principle and so he is “frankly, getting a little tired of the topic.” I take it then that there a substantial literature on this topic (unless all this took place in private conversations), but Fodor in his breezy style doesn’t bother to cite it.

    My own sense is that the question in general and in the particular, important case of causing harm is going to turn on our analysis of causation, as some suitably strong sense of cause is wanted to get the biconditional to go through. One strategy Fodor dismisses – replacing “cause” with “immediately cause” or some locution like that – actually seems pretty promising to me, although I’ll have to think more about Pavel Davydov’s suggestion about the “being prominent in the causal ancestry of” relation. If we take the “immediately” seriously, then no other agents can intervene in the chain of causation. Ball A striking ball B which strikes ball C is clearly a cause of ball A causing ball C to move, but when we switch to agents, making this attribution work suddenly becomes much harder if not impossible.

Leave a Reply