I am very pleased to contribute to this symposium on Causation with a Human Face (CHF). My commentary concerns chapter 6 of CHF, which uses the notion of invariance to shed light on various puzzle cases about causation. (See Jim’s post “Invariance and Distinctions Within Causation” for a summary of the chapter.) I will focus on Jim’s discussion of causation by absence.
One question raised by Jim’s discussion concerns the role (if any) of normality in explaining causal judgments regarding absences. Jim acknowledges in chapter 2 of CHF that considerations of abnormality (encompassing violations of statistical, moral and functional norms) play an important role in causal selection, and also grants that issues of moral responsibility and which possibilities we take seriously may play a role in explaining at least some of our causal judgments about absences (p. 292). Invariance considerations are supposed to constitute an additional influence on those judgments. But one may wonder whether normality considerations alone are sufficient to explain our assessments, without having to appeal to invariance. Consider a simple account (in the spirit of McGrath 2005) on which only abnormal absences are regarded as causes. That story makes sense of the examples discussed by Jim: we do not think that the absence of a meteor hitting Earth (normal) caused Jim to write his paragraph, but we do regard the absence of food (abnormal) as a cause of the death. (Here I leave aside reasons one may have to be skeptical of proposals that give such a direct role to normality considerations in causal cognition: see Blanchard and Schaffer 2017.) It would be interesting to hear how Jim himself views the respective contributions of normality and invariance in explaining causal judgments involving absences.
This issue, note, is intimately related to the question of the function(s) of actual cause judgments. One idea is that actual cause judgments seek to identify dependence relationships that are exportable to other circumstances and hence useful for future prediction and control (Lombrozo and Carey 2006; Hitchcock 2012). On that hypothesis, we should naturally expect actual cause judgments (including judgments about absences) to favor factors that are highly invariant under changes in background conditions. On the other hand, as far as I can see, the proposal does not predict a strong influence of normality considerations in actual cause judgments. A somewhat different hypothesis is that the goal of actual cause judgments is generally to identify appropriate targets of interventions that would alter the effect (Hitchcock and Knobe 2009). That proposal does give a central place to normality considerations in actual cause assessments, as abnormal factors are generally easier and more appropriate targets of effect-changing interventions. So disentangling the respective roles of normality vs. invariance considerations might contribute to shed further light on the question of what actual cause judgments are for. (Though one wrinkle here is that Hitchcock and Knobe’s hypothesis may in fact predict a preference for invariant relationships too. Suppose that c is a highly invariant cause of e. Then other circumstances besides c will be poor targets of interventions for changing e, for c’s invariance entails that e would still happen if those circumstances were varied. Provided that changing c would change e, we may therefore expect people to choose c when asked what caused e.)
Besides invariance, another “distinction within causation” that Jim discusses in CHF is influence (in Lewis’s sense), which concerns the extent to which the time and manner of occurrence of the effect depends on the time and manner of occurrence of the cause. Invariance and influence are tightly related to one another, as Jim notes in the book, but they can nonetheless come apart. A causal relationship may obtain only in rarefied circumstances, and yet be such that once those circumstances are in place the cause has a high degree of influence over the effect. There is considerable evidence (much of it adduced by Jim himself, see e.g. Woodward 2010) that influence plays an important role in various aspects of causal thinking. Moreover there is a strong normative rationale for caring about causes that score high on influence since they enable fine-grained control of their effects. And interestingly, in the examples discussed by Jim those absences that seem more paradigmatically causal also have more influence on their effects. Not being hit by a meteor has very little influence on the content of Jim’s paragraph, whereas withdrawing food has a large fair amount of influence on the victim’s death. (The time and manner in which food is withdrawn will affect the time of the death, how long it takes, etc.) So one may wonder if and to what extent our responses to causal claims about absences might be explained by influence considerations rather than – or, more plausibly, in addition to – invariance.
Of course, in the end it is an empirical question how much invariance, normality and influence each contribute to explaining our causal judgments about absences. So my remarks only reinforce one of Jim’s points in chapter 6, namely that more systematic empirical exploration of the relevant issues would be welcome.
Thomas Blanchard (University of Cologne)
tblancha@uni-koeln.de
References
Blanchard, T. & Schaffer, J. (2017). Cause without Default. In Beebee, H., Hitchcock, C. & Price, H. (Eds.) Making a Difference: Essays on the Philosophy of Causation, pp. 175-214. Oxford: Oxford University Press.
Hitchcock, C. (2012). Portable Causal Dependence: A Tale of Consilience. Philosophy of Science, 79: 942-51.
Hitchcock, C. & Knobe, J. (2009). Cause and Norm. Journal of Philosophy, 106: 587-612.
Lombrozo, T. & Carey, S. (2006). Functional Explanation and the Function of Explanation. Cognition,99: 167-204.
McGrath, S. (2005). Causation by Omission: A Dilemma. Philosophical Studies, 123: 125-48.
Woodward, J. (2010). Causation in Biology: Stability, Specificity, and the Choice of Levels of Explanation. Biology & Philosophy, 25: 287-318.
Featured image via John Chumack
Response to Thomas Blanchard
I thank Thomas Blanchard for his very rich and interesting commentary. For reasons of space I will not try to respond to all of his questions and suggestions here, but I will take up several of them. Before doing that, though, I also want to say that like the other participants in this symposium, he has also been able to combine both “philosophical” work on causation and causal reasoning in highly illuminating ways, including joint work on the role of invariance in causal judgment (the Vasilyeva et al., 2018 paper mentioned in earlier an earlier post) and empirical work on how people in fact judge in causal exclusion problems– they don’t judge in the way advocated by fans of the exclusion argument (Blanchard et al. forthcoming). As explained in an earlier post, when we have both normative arguments supporting the appropriateness of certain judgments (in this case, judgments that do not conform to exclusionist assumptions, as I have argued in CHF and elsewhere) and empirical results that accord with these normative arguments, the combined impact of these two sorts of considerations can be particularly compelling.
As Thomas notes, CHF acknowledges a role for “normality” in causal judgments. He raises the question of whether normality considerations alone without appeal to invariance are sufficient to explain causal judgments in many of the cases, including the case of causation by absence, that I discuss. (As he explains, the idea here is that abnormal causes are more readily cited as causes than normal ones.) As Thomas observes, invariance and normality considerations are often confounded in particular cases, which can make it difficult to separate their influences. To use an example from Thomas (and CHF), (1) “My failure to be struck by a large meteor caused me to write these very words” strikes us as (at least) non-standard or non-paradigmatic. However, (1) is both relatively non-invariant and involves a candidate cause that is very normal, so both considerations are present and point in the same direction– that is, in support of the claim that (1) is non-standard.
One way of disentangling the role of these two considerations is to consider scenarios (ideally presented in experiments, rather than just considered from the armchair) in which the considerations are manipulated separately. For example, the relation between (2) complete deprivation of food (F) and death (D) is highly invariant and Thomas suggests that F is also abnormal, so that both considerations support the claim that (2) is acceptable. But, assuming that the frequency with which F occurs has some influence on whether it is regarded as normal, we can manipulate this frequency while holding the F–> D relation fixed. That is, we can consider cases in which deprivation of food is very common, as it was, for example, in 1930s Stalinist Russia. Would people be less likely to judge that F causes death in such cases? If, as I suspect, the answer is “no”, this would show that the invariance of the F–> D relation has an independent effect on our judgments. (I’m glossing over a number of complexities here; for example, as Icard et al. 2017 — discussed previously– show, in cases of causal selection not involving absences, normality of the cause does influence causal selection judgments –see below– although in ways that also depend on the nature of the functional form connecting the cause and the effect.)
Thomas also notes that considerations about the relative role of invariance and normality have interesting implications concerning the function of actual cause judgments. He suggests that if we think of actual cause judgments as having the function of identifying dependence relationships that are exportable to other circumstances and hence useful for future prediction and control, then we would expect such judgments to be strongly influenced by invariance considerations, but not so much by normality considerations. On the other hand , if the function is to identify appropriate targets of interventions that would alter the effect, this would fit nicely with a role for normality in causal judgments.
I fully agree with this analysis. Indeed, I would praise it as brilliant and highly insightful were it not for the fact that in a forthcoming paper (Woodward, forthcoming) which neither Thomas or most of the other commentators have seen, I adopt something very like it myself. I tentatively argue there that what are called actual cause judgments in the philosophical literature are a heterogeneous lot, with at least two different functions– the two mentioned by Thomas above– and, correspondingly, two different standards of assessment. (Since, recall, we assess correctness or appropriateness on the basis of function.) One kind of judgment often has to do with causal selection in the sense that we select one candidate as “the cause” from several factors that are thought to be causally relevant to an effect –the dropped match rather than the presence of oxygen is selected as the cause of the fire. Considerations of normality strongly influence such selection practices and there is an obvious rationale for this in terms of which factor is the most appropriate target of intervention (the match, not oxygen). In other cases, though, including cases which are not so naturally regarded as having to do with causal selection, considerations of invariance seem to play a more prominent role in causal judgment. This is the case, for example, in standard causal pre-emption and over-determination scenarios. So my answer to Thomas’ question about the role of invariance and normality in actual cause judgment is that both play a role but have different weights for different sorts of actual cause judgments. I suggest this may be one reason why it has proved so difficult to come up with a single unitary theory of such judgments.
Blanchard, T., Murray, D. & Lombrozo, T. “Experiments on Causal Exclusion.” Mind & Language forthcoming.
Woodward, J. (Forthcoming) “Actual Causation: It’s Complicated” In Advances in Experimental Philosophy of Causation (ed. Willemsen and Wiegmann) Bloomsbury Press.