Abstract
Recently, there has been a large amount of support for the idea that causal claims can be sensitive to normative considerations. Previous work has focused on the concept of actual causation, defending the claim that whether or not some token event c is a cause of another token event e is influenced by both statistical and prescriptive norms. I focus on the policy debate surrounding alternative energies, and use the causal modelling framework to show that in this context, people’s normative commitments don’t just affect the causal claims they are willing to endorse, but also their understanding of the causal structure. In the context of the alternative energy debate, normative considerations affect our (implicit) understanding of the causal structure by influencing our judgements about which variables should be held fixed, and therefore which variables should be relegated to the background of a causal model. In cases of extreme disagreement, normative commitments can also affect which causal structure we think should be instantiated. Thus, focusing on a new context has revealed a previously unexplored sense in which normative factors are incorporated into causal reasoning.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Of course, there is a sense in which the causal structure remains the same before and after such an intervention, and it is generally possible to represent both situations using a single, larger, causal model. The point is that when the intervention range is exceeded, the causal structure represented by the original causal model is no longer instantiated. There is therefore a sense in which the causal structure has changed.
Questions about which causal model is appropriate should not be understood as pertaining to the metaphysics of causation, but to a different, methodological, question about how we should go about causal reasoning. For discussion, see Woodward (2015).
It has previously been observed that normative factors make a difference to causal judgements in the sense that we judge x to be a cause of y if y counterfactually depends on xrelative to a default situation (Hall 2007; Hitchcock 2007; Halpern and Hitchcock 2010). However, the examples used to illustrate this phenomenon tend to start by representing the relevant situation using a particular causal structure, and then show that in order to account for our causal judgements, we have to add default values to this structure, where the default values are determined by both statistical and prescriptive norms. What I emphasise is that normative factors can (legitimately) influence the first step—that is, what we take to be the appropriate causal structure itself. As discussed in footnote 13, this can be seen as extending a suggestion from Blanchard and Schaffer (2017).
This approach originates in Hart and Honoré (1959). For recent statements, see the works cited in Footnote 1.
Hitchcock and Knobe point out that it is the concept of actual causation that has been the primary target of interest in the philosophy of causation (2009, p. 587). Judgements of actual causation can be understood as answers to the question: which event(s) was causally responsible for a particular effect? See Woodward (2011).
An obvious objection to the kind of account described above is that it confuses causation with blame. Perhaps normative considerations only enter into our causal judgements to the extent that we are failing to distinguish causation and blame. Empirical studies have shown that judgements of actual causation are affected by negative evaluations such as blame (see e.g. Alicke et al. 2011). However, this doesn’t exhaust the role played by the normative. Other studies have shown that we are more likely to cite deviations from the norm as actual causes, even when the outcome is positive (see e.g. Hitchcock and Knobe 2009).
See also Halpern and Hitchcock (2015).
Halpern and Hitchcock consider examples in which there is disagreement about the relevant norms in their (2015).
Perhaps it is more likely that there was general agreement about which norms held, but that different people prioritised these differently. This would also explain the disagreement over the cause of the riots.
One way of understanding this project is as an elaboration of an idea from Blanchard and Schaffer (2017). In opposition to the work described above, they argue that we don’t need to incorporate the deviant/default distinction into our account of actual causation. Rather, we need to pay more attention to what it takes to be an apt causal model. It is an implication of their argument (which they don’t discuss) that there will be cases in which normative commitments influence which causal model we consider to be appropriate—that is, our understanding of the causal structure.
Roughly, an intervention variable on X with respect to Y has to be a cause of X, and has to affect Y (if at all) only via X. A random controlled trial is therefore a paradigmatic example of an intervention: the whole point of this experimental design is to ensure (as best as possible) that confounding factors are controlled for—that is, that any effect on the dependent variable is due to the independent variable. For Woodward’s precise characterisation of the notion of an intervention, see 2003, pp. 98–99.
Notice that causal models encode a set of counterfactuals. For example, Fig. 1 asserts that there is a possible intervention on the amount of rainfall (R) that makes a difference to the river level (RL). This entails that there is a true counterfactual with the following form: if it were to rain x amount (rather than x’ amount), the river level would be y (rather than y’).
Thanks to an anonymous reviewer for pressing me to clarify the role played by the causal modelling framework in my argument.
A clarification is required here. It is not the case that all variables that are in the background (as opposed to the foreground) should be held fixed. In particular, background variables that are causally intermediate between a cause and its effects should be allowed to vary freely. For example, consider Fig. 1. We could move the variable RL into the background, and represent R, V, and S as direct causes of F. In this scenario, if we were to hold RL fixed, then intervening on any of R, V, or S would not make a difference to the value of F, and so these variables would not come out as causally relevant. Thus, if the model under consideration is to correctly represent the causal relationships between these four variables, we had better allow RL to vary. More precisely, then, the requirement is that background variables that are not on a causal pathway between any of the foreground variables are held fixed at some value.
Note that the background/foreground distinction differs from the normal/deviant distinction in that the former holds between variables, whereas the latter holds between values of variables. For example, imagine waking up tomorrow morning to discover that the earth’s gravitational field has become much stronger overnight. This would represent a deviant value of the variable {magnitude of the earth’s gravitational field}. We would probably respond by (either implicitly or explicitly) moving this variable from the background to the foreground of many causal models, including Fig. 1.
For my purposes, it is not important to give a precise definition of the term ‘renewables’. However, assume that it refers primarily to wind and solar.
Another way of putting this point is to say that the two individuals discussed above disagree about the utility values that should be assigned to the values of the variables {employment level} and {air pollution}, and therefore how much these contribute to standard of living.
There are also likely to be situations in which two individuals both agree that it is possible to convert to renewables while (roughly) maintaining our current standard of living, but disagree about whether this is desirable. See Sect. 6 for further discussion.
Two clarifications. First, I leave aside the question of exactly what it would take to ‘convert to renewables’. Perhaps generating 90% of a country’s energy requirements using renewables would fulfil this criterion, or perhaps not. I suspect that different people have different targets in mind. Second, there is also a distributional issue. It may turn out that it is possible to convert to renewables without any reduction in a country’s GDP, but that the conversion nevertheless has unacceptable consequences for certain groups of people. I also set this issue aside for the purposes of this paper.
Strictly speaking, I do not provide causal models, but causal graphs. To turn these into causal models, you would need to add a set of structural equations.
ProCon.org (2017a) This site cites 19 arguments for, and 18 against, the above question.
ProCon.org (2017b) This cites 7 arguments both for and against government subsidies of alternative energies. I included the responses to this second question because government subsidies came up in response to the question ‘Can alternative energy effectively replace fossil fuels?’, but were addressed more comprehensively in response to the more specific question.
I assume that for it to be possible to convert to renewables is for there to be some practically realistic intervention (or set of interventions) that would result in this conversion. In terms of the causal models discussed below, this restricts us to considering interventions on the exogenous variables.
To dispel any worries that \(EH_{F}\) and \(EH_{R}\), as characterised above, are not apt for including in a causal model, notice that they are distinct from the other variables in the model, capable of being intervened on, and that we can assign them a set of possible values that represent incompatible states of affairs.
It may seem strange to have a yes/no variable as an effect in this model. An alternative way of understanding this variable is as follows: Assume that the other variables in the model all represent the situation at some time t = 0. Y/N can then be understood as representing the maximum proportion of energy that could be produced by renewables at some later time t = 1.
Some advocates of fossil fuels argue that the government should only subsidise renewables if these are close to being competitive with fossil fuels on price. That is, they argue that there should be a causal arrow going from \(P_{R}\) to \(GS_{R}\) (or perhaps from Y/N to \(GS_{R}\)). I leave out this postulated causal link.
For a review of work on motivational bias, see Kunda (1990).
Thanks to two anonymous reviewers for pressing this objection.
It is possible that some of the apparent differences in people’s understanding of the causal structure are actually due to rhetoric—that is, that people on opposing sides of the debate agree on the causal structure, but only emphasise those aspects of the structure that support their argument. However, it seems implausible that rhetoric can fully account for the differences represented in Fig. 3.
This first question may appear to be empirical; certainly it is partly empirical, and it is possible to operationalise environmental and health costs such that they can be assessed empirically. However, it seems unlikely that there is a value neutral choice of operationalisation. If not, there is a genuinely normative question here.
Another way of putting this point is to note that if you think that a variable shouldn’t be intervened on, you think that there is only one value of that variable that represents a serious possibility. You therefore don’t need to represent other values of this variable in your causal model. For further discussion, see (2017, pp. 197–198).
Whether or not we think that the government should subsidise renewables and/or fossil fuels (and by how much) is another place in which there is a genuine difference in normative commitments that affects which variables are in the foreground and which are in the background.
Note that normative factors can also make a difference to our assessment of the value of certain variables (e.g. environmental and health costs). Incorporating all the variables that are considered to be suitable targets of intervention by some people engaged in the relevant debate will also fail to resolve this kind of disagreement.
The economy is not assumed to be constant in the sense that any specified variable (or set of variables) is held fixed. Rather, it is assumed that we don’t exceed the invariance range of the graph in Fig. 2.
The causal structure of a conserver economy may be quite different in different locations—even between different cities in the same country.
The kind of intervention described above is often referred to as a structural intervention. For example, see Daniel (2017). The structural interventions Malinsky considers involve changes to the parameters of structural equations. However, there are also situations in which we can affect the causal structure more dramatically. For example, some manipulations change which variables are causally relevant to a given effect (as described above). In other work, I have argued that this is actually a very common practice, which is essential to our ability to control the world (2017).
See footnote 15 on the connection between causal structure and counterfactuals.
For more discussion, see Halpern and Hitchcock (2010).
References
Alicke, M. D., Rose, D., & Bloom, D. (2011). Causation, norm violation, and culpable control. Journal of Philosophy, 108, 670–696.
Blanchard, T., & Schaffer, J. (2017). Cause without default. In H. Beebee, C. Hitchcock, & H. Price (Eds.), Making a difference: Essays on the philosophy of causation (pp. 175–214). Oxford: Oxford University Press.
Gopnik, A., & Schulz, L. (Eds.). (2007). Causal learning: Psychology, philosophy, and computation. Oxford: Oxford University Press.
Hall, N. (2007). Structural equations and causation. Philosophical Studies, 132, 109–136.
Halpern, J., & Hitchcock, C. (2010). Actual causation and the art of modelling. In R. Dechter, H. Geffner, & J. Y. Halpern (Eds.), Probability and causality: A tribute to Judea Pearl (pp. 383–406). London: College Publications.
Halpern, J. Y., & Hitchcock, C. (2015). Graded causation and defaults. British Journal for the Philosophy of Science, 66, 413–457.
Halpern, J. Y., & Pearl, J. (2005). Causes and explanations: A structural-model approach. Part 1: Causes. British Journal for the Philosophy of Science, 56, 843–887.
Hart, H. L. A., & Honoré, T. (1959). Causation in the law. Oxford: Clarendon Press.
Hitchcock, C. (2007). Prevention, preemption, and the principle of sufficient reason. Philosophical Review, 116, 495–532.
Hitchcock, C., & Knobe, J. (2009). Cause and norm. Journal of Philosophy, 106, 587–612.
Ismael, J. (2013). Causation, free will, and naturalism. In D. Ross, J. Ladyman, & H. Kincaid (Eds.), Scientific metaphysics (pp. 208–235). Oxford: Oxford University Press.
Knobe, J. (2010). Person as scientist, person as moralist. Behavioral and Brain Sciences, 33, 315–329.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108, 480–498.
Lagnado, D. (2011). Causal thinking. In P. M. K. Illari, F. Russo, & J. Williamson (Eds.), Causality in the sciences (pp. 129–149). Oxford: Oxford University Press.
Malinsky, D. (2017). Intervening on structure. Synthese. https://doi.org/10.1007/s11229-017-1341-z.
McGrath, S. (2005). Causation by omission: A dilemma. Philosophical Studies, 123, 125–148.
Menzies, P. (2004). Difference-making in context. In J. Collins, N. Hall, & L. A. Paul (Eds.), Causation and counterfactuals (pp. 139–180). Cambridge, MA: MIT Press.
Menzies, P. (2007). Causation in context. In H. Price & R. Corry (Eds.), Causation (pp. 191–223)., Physics, and the constitution of reality Oxford: Clarendon Press.
Menzies, P. (2009). Platitudes and counterexamples. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation (pp. 341–367). Oxford: Oxford University Press.
ProCon.org. (2017a). Can alternative energy effectively replace fossil fuels? http://alternativeenergy.procon.org/view.answers.php?questionID=001244.
ProCon.org. (2017b). Should the us subsidize alternative energies? http://alternativeenergy.procon.org/view.answers.php?questionID=001251.
Statham, G. (2017). The manipulation of chemical reactions: Probing the limits of interventionism. Synthese, 194, 4815–4838.
Systma, J., Livengood, J., & Rose, D. (2012). Two types of typicality: Rethinking the role of statistical typicality in ordinary causal attributions. Studies in History and Philosophy of Science Part C, 43, 814–820.
Woodward, J. (2003). Making things happen: A theory of causal explanation. Oxford: Oxford University Press.
Woodward, J. (2011). Psychological studies of causal and counterfactual reasoning. In C. Hoerl, T. McCormack, & S. R. Beck (Eds.), Understanding counterfactuals (pp. 16–53)., Understanding causation Oxford: Oxford University Press.
Woodward, J. (2015). Methodology, ontology, and interventionism. Synthese, 192, 3577–3599.
Woodward, J. (2016). The problem of variable choice. Synthese, 193, 1047–1072.
Acknowledgements
Thank you to Claire Benn, Sharon Berry, Casper Storm-Hansen, Moshe Justman, York Hagmayer, Christopher Hitchcock, Enno Fischer, participants at the Linguistic Perspectives on Causation Workshop at the Language Logic and Cognition Center, Hebrew University of Jerusalem (June 2017) and the Polonsky Academy Seminar (May 2017), and especially to three anonymous reviewers from this journal.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Statham, G. Normative commitments, causal structure, and policy disagreement. Synthese 197, 1983–2003 (2020). https://doi.org/10.1007/s11229-018-1775-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-018-1775-y