In this paper, I argue that we should adhere to manipulationism as the only account of causation in the social sciences, since it suffices to explain all causal relations in the social world (i.e. no pluralism of causal accounts is needed). I start by outlining manipulationism. Then I put forward an evaluative criterion any exhaustive account of causation should meet (the “desideratum absolutum”). After that, I present Prescott-Couch’s (2017) case against manipulationism. He makes two objections against the manipulationist account, one based on the existence of “Perceived Abnormality Cases” and one based on the existence of “Ontological Dependence Cases”. The former supposedly demonstrate instances where relations are manipulable yet not causal, the latter instances where relations are causal yet not manipulable. I show that the manipulationist can use social scientific laws to calculate and subtract the perceived abnormalities in the former through the use of social scientific laws, and thus manipulationism can account for perceived abnormality cases. I then lay out how ontological dependence cases can be transformed into perceived abnormality cases, and, once that is done, how the manipulationist can use the same social scientific laws to account for them. Since we have not yet encountered a category of social phenomena manipulationism cannot account for, I conclude that manipulationism should be upheld as the only causal paradigm in the social sciences.
The manipulationist account of causation
The manipulationist account is a counterfactual account of causation, meaning that it presupposes a dependence relationship between cause (A) and effect (B) (Menzies & Beebee, 2019). When we say that A and B are related through dependence, we stipulate that, had A not occurred, B would not have occurred. Manipulationism goes further than many counterfactual accounts of causation, for it does not only assert that B would not occur if A had not occurred, but that if one were to intervene on A, B would necessarily change in accordance to that intervention. Hence, for the manipulationist, A is a total cause of B if and only if (iff) under an intervention that changes the value of A (other variables held constant) an associated change in the value of B occurs. It is important to note that the intervention must be both isolated and exogenous. An intervention is isolated iff it changes the value of A without altering any other values relevant to the causal structure in which A and B are embedded. And an intervention is exogenous iff A, in the process of the intervention, is cut off from parts of the causal structure that would otherwise determine its value. As these two strict conditions must be satisfied, we refer to any such intervention as a “surgical intervention”.
The Desideratum Absolutum
Before examining the first argument against manipulationism, I present my criterion for assessing any account’s suitability as the causal paradigm in the social sciences. I call my criterion the “desideratum absolutum”. This criterion stipulates that an account of causation C can take a paradigmatic role in the social sciences iff there exists no set of social phenomena P whose causal structures cannot be explained by C. I use the term “paradigmatic role” in a Kuhnian sense; an account is considered to carry a paradigmatic role iff it is universally recognized and applied throughout the discipline concerned (in this instance, the social sciences) (Kuhn, 1996, p. 10). For an account of causation, this would mean that social scientists universally use it to explain causal relations. A social phenomenon’s causal structure is not explained by C iff C misses causal relations that are present in the causal structure or C adds causal relations not present in the causal structure.
Using my desideratum sets a very high bar to manipulationism. If it fails to explain any category of social phenomena, it fails as a causal paradigm. I justify this with the argument that an account which misexplains a part of the social world clearly cannot explain the social world alone. At the very least, the category of social phenomena it cannot explain would need to be explained by a different account.
Note that it is theoretically possible for several accounts of causation to fulfil this desideratum. However, if we can show that manipulationism fulfils it, then there is no need to make use of other such accounts, because causation in the social world would already be fully accounted for.
One could, quite reasonably, contest my desideratum because it places such strict limitations on what can constitute a “causal paradigm”. Concerning an account of causation, one could argue that social scientists should adhere to it as long as it correctly explains most social phenomena’s causal structure. After all, in modern scientific practice, many theories are not discarded as soon as they misexplain a certain class of phenomena. I concede that such concerns are valid. Yet I favour such a strict desideratum because I believe manipulationism can meet its standard. And by doing so, manipulationism allows us to make an even stronger case for its use in the social sciences.
The Perceived Abnormality Cases
I now discuss the first argument against manipulationism, which Prescott-Couch (2017) made in form of the “perceived abnormality cases” (pp. 491-502). Prescott-Couch argues that these cases exemplify instances where relations are manipulable (i.e. surgical intervention is possible), but there is no causal relationship. As a result, manipulationism fails to explain their causal structure. To illustrate his argument, Prescott-Couch (2017) presents us with the case of George, who is “a child from a single parent household [that] has developed serious distress. One factor influencing his distress is lacking contact with one parent and another is facing social stigma attaching to children of single-parent homes.” (p. 494). Put differently, there are two causal relations at work. Firstly, being in a single-parent household (P=1) has a direct effect on George’s level of distress. Secondly, being in a single-parent household leads to stigma (S=1), which then affects his level of distress.
Figure 1, (p. 494)
Now, performing a surgical intervention, we hold S fixed at 1 and set P to 0.
Now, performing a surgical intervention, we hold S fixed at 1 and set P to 0. Consequently, George experiences a higher level of distress than when P=1, because now he lives out a Kafkaesque nightmare, where he has two parents, yet is taunted for being from a single-parent household. That is, in Prescott-Couch’s terms, a case of “perceived abnormality”, because the fact that George has only one parent clearly appears to be something that contributes to distress, yet, under manipulationism, him having one parent actually prevents distress. Therefore, as manipulationism asserts that George’s distress goes up when P is set to 0, it misidentifies the causal contribution P makes to D. The relation between P and D is hence, while manipulable, not causal. And as manipulationism conflates these two notions, causal structures resulting in such perceived abnormality cannot be explained by it.
While intuitively appealing, I believe there is a persuasive way to defend manipulationism against the perceived abnormality cases. That way consists in manipulating perceived abnormality cases through a lens of metaphysical possibility (Woodward, 2004, pp. 50-53). A manipulationist can carry out a surgical intervention in a logical sense even if existing conditions would not allow it to occur in the real world. The example Woodward (2004) gives in Making Things Happen is that of a scientist who wishes to investigate what effect the position of the moon has on the tides (pp. 52-53). The scientist posits the question of what would occur if the moon were 338,900 miles away from the earth instead of 238,900. Unfortunately for the scientist, there appears to be no way in which an appropriate intervention could be carried out. Were we, for instance, to change the moon’s position through changing the position of another large object X, this would violate manipulationism, as this change would itself exert a gravitational effect on the tides. The intervention would hence not be surgical, as it would not be isolated. Clearly, there exists no physically possible way to change the position of the moon without changing the values of other variables impacting the tides. One can, however, acknowledge this reality without abandoning manipulationism. The way to do this is through, hypothetically, moving the position of the moon and then removing the values of confounders through calculation. If we take our example of X that moved the moon’s position, this would mean that we calculate its effect on the tides through Newtonian physics and then subtract it from the value of the tides. This then leaves us with the effect that moving the position of the moon has on the tides.
The important question now is whether manipulation by metaphysical possibility, as opposed to practical possibility, can rescue the category of “perceived abnormality cases” in the social world from Prescott-Couch’s critique. In order to demonstrate that it can, I show that the causal structure of George’s case can be captured through a manipulationist account employing metaphysical possibility. With George’s case, we face the same fundamental problem as with the moon’s position: like the effect of X on the tides, we can think of the stigma in the causal structure of George’s case as a confounder of the causal relation we want to observe. With the former, it was possible to employ Newtonian laws to subtract the confounding influence of X. Sadly, we cannot employ Newtonian physics to calculate the value of a stigma. Like most social phenomena, stigmas are constituted by multiple weak causes, all of which are sensitive to interference, and, on top of that, embedded into nested socials structures with complex interdependences.
Admittedly, one cannot credibly deny that the nature of such social phenomena poses a challenge for the manipulationist account. However, I hold that this is not a problem created by manipulationism, but rather something that is intrinsic to the social world. The complexity of the social world is something that all social sciences struggle with, which is why laws describing causal regularities in social science are prone to more exceptions and less generalisability.
Yet, even though the causal regularities in the social world can be considered less stable, is one willing to go as far as to say that the social world has no causal regularities, or that these causal regularities cannot be described? If the answer is yes, then one’s problem is not with manipulationism, but with the social scientific endeavour. If the answer is no, then one will have to concede that we can, in principle, determine the causal influence an event X has on event Y through some law Z, where the law Z is a description of the causal regularity.note1
For George’s case, that means that it is possible for us to estimate the impact that stigma has on his distress. For that, we will, quite likely, have to make reference to a psychological law that relates the effect stigmatisation takes on one’s mental health. By doing so, we can then subtract the effect of the stigma from his level of distress. This leaves us with the causal influence of being from a single-parent household on his distress, or, put differently, his level of distress is brought back to 0. Thus, we have captured the causal structure of George’s case despite a ‘perceived abnormality’.
Eventually, we have good reason to believe that, by considering metaphysical possibilities for intervention rather than practical ones, the manipulationist can account for all perceived abnormality cases. That is because, as long as one is not willing to deny that the social world has describable causal regularities, one ought to acknowledge that the confounding influence a variable exerts on another can be principally removed. And just as we demonstrated with George’s case, this then makes it possible for manipulationism to explain that case’s causal structure accurately. Therefore, manipulationism survives this objection and meets our desideratum absolutum.
The Ontological Dependence Cases
I go over to discussing Prescott-Couch’s (2017) second argument, presented in form of the “ontological dependence cases” (pp. 502-510). These cases try to show that there exist social phenomena whose relations are causal yet not manipulable (i.e. intervention is not possible). As a result, manipulationism fails to explain their causal structure. To illustrate his argument, Prescott-Couch presents us with the case of female priesthood in the Episcopal church: “Since the Episcopal church allowed women to enter the priesthood, female congregants in the Church of the Incarnation have felt more satisfied in church for two reasons: because of changes to the institution of the priesthood and because of their interactions with their new female priest” (p. 502). Put differently, there are two causal relations at work. Firstly, the institutional rule allowing female priests (R=1) has a direct effect on the satisfaction level of female congregation members. Secondly, the institutional rule allows for the presence and activity of female priests (F=1), which then affects the level of satisfaction among females of the congregation (H).
Figure 2, (p. 502)
Now, performing a surgical intervention, we hold F fixed at 1 and set R to 0.
Immediately, we face a grave problem for the manipulationist account. That is because it is metaphysically impossible for female priests to be active if the institutional rules do not allow for females to join the priesthood. Such metaphysical impossibilities stem from the way in which the social world is constructed, for individuals rely upon institutional recognition to be acknowledged in a social role. A woman, for example, is only a priest in virtue of her church recognising her as such, which is why females in the Episcopal Church could not be active as priestesses if the Episcopal Church did not allow for them to practise as priests. Therefore, one cannot perform a surgical intervention on F whilst holding R fixed, because the relation between F and R cannot be isolated from H – a variable which is clearly relevant to the causal structure of the case. Manipulationism hence falsely holds that there exists no causal relationship between R and H.
A successful manipulationist defence, I think, lies in rejecting that something like “ontological dependence cases” exist in the social world. This may appear to be a counter-intuitive line of reasoning, but I believe it is ultimately successful.
I substantiate it with the claim that, for any ontological dependence case O, there exists no background condition X which exerts an effect Z on a social phenomenon through a different route than a change in their beliefs Y. Put differently, a background condition only exerts causal influence on social phenomena through affecting the beliefs of people. Whatever social phenomenon we think of, the causal effect of “background conditions” (be they about race, gender, income, etc.) ultimately boil down to affecting the beliefs members of society hold. For instance, if institutionalised gender bias is taken to be a factor in job applications, then surely one will acknowledge that this institutionalised gender bias only exerts an effect on an applicants’ chances in virtue of how employers think about the two genders. It is entirely unclear how the institutionalised gender bias, or any other background condition, for that matter, could exert a causal influence otherwise. As long as there is no case to prove the contrary, we hence maintain that background conditions only carry causal impact through affecting people’s beliefs.
For that reason, when one analyses any “ontological dependence” case, one will realise that such a thing as ontological dependence does not actually exist in the social world, for an intervening variable stands between any background condition and social phenomenon. I illustrate this with the following reconstruction of Prescott-Couch’s (2017) priesthood case:
Once we install female congregation members’ belief on whether women can join the priesthood (R2) as a binary variable in the model, it is clearly not the institutional rule (here the background condition) that affects the female members’ satisfaction, but their belief about the institutional rule itself that exerts this influence. In this model, R=1, i.e. the institutional rule allowing for female priests, causes R2=1, i.e. female congregation members believing that women can join the priesthood, which then causes an increase in H, i.e. the female congregation members’ happiness. I regard the insertion of R2 in the model as uncontroversial, for clearly it is logically possible that the Episcopal Church could allow females to join the priesthood, but congregation members could still think that female members are not allowed to join the priesthood. The addition of R2 hence does not change the priesthood case but uncovers an additional aspect of its causal structure.
This distinction is nonetheless significant, as it reveals that Prescott-Couch’s “ontological dependence case” is, in reality, just another “perceived abnormality case”. Of course, it appears abnormal for female congregation members to think that they cannot join the priesthood when they, in fact, can do so. Similar to George’s case, we find ourselves in a Kafkaesque situation, where an increase in female congregation members’ happiness is prevented because they believe there can be no female priests, when in fact, the Church is making it clear that women can join the priesthood. But as demonstrated in our response to the perceived abnormality cases, intervention in such apparently abnormal cases is metaphysically possible. To demonstrate this, let us set R2 to 0 (other variables held constant). Recall that R2 displays the belief of female congregation members about the institutional rule on women joining the priesthood. Setting R2 to 0 hence means that female congregation members do not believe females can join the priesthood. That this is a perceived abnormality does not matter, because it is, in principle, possible to subtract the effect that R2 has on H, which then leaves us with the causal effect of F on H. Again, like in George’s case, one will not be able to use Newtonian laws to remove the confounding effect of R2. We would once again need to employ a potentially less reliable and less generalisable social scientific law, such as a psychological law relating social representation to happiness, to estimate the effect of R2 on H.
Yet as long as one is not willing to claim that the social world is one without describable causal regularities, one will have to acknowledge that it is, in principle, possible to remove the confounding influence in the priest case. This is significant for all ontological dependence cases, since manipulationism can, as demonstrated, account for any background conditions through further uncovering such a case’s causal structure. Hence, ontological dependence cases do not violate the desideratum absolutum.
I refuted Prescott-Couch’s first objection by showing that accounting for “perceived abnormalities” is metaphysically possible, and his second one through demonstrating that ontological dependence cases can be turned into perceived abnormality cases and hence accounted for in the same way. Thus far, manipulationism has not failed to account for even one category of social phenomena, so we can confidently infer that it meets our strict desideratum for a causal account. This is, of course, not to say that it is impossible for there to exist a category of social phenomena which manipulationism cannot account for. However, as long as we have not encountered any such category, causal pluralism is superfluous. As manipulationism suffices to explain causal relations in the social world, it should be upheld as the only account of causation in the social sciences.
Some philosophers may shy away from using the term ‘law’ in the context of social scientific research, but as I use the term in the broadest sense, namely as the description of a causal regularity, raising this as an issue would be nothing more than a semantic objection.
- Kuhn, T., 1996. The Structure of Scientific Revolution. s.l.:University of Chicago Press.
- Menzies, P. & Beebee, H., 2019. Counterfactual Theories of Causation. [online] Available at: https://plato.stanford.edu/entries/causation-counterfactual/ [Accessed 14 11 2020].
- Prescott-Couch, A., 2017. Explanation and Manipulation. In: s.l.:NOUS 51:3, p. 484–520.
- Woodward, J., 2004. Making Things Happen: A Theory of Causal Explanation.. [Online] Available at: https://oxford.universitypressscholarship.com/view/10.1093/0195155270.001.0001/acprof-9780195155273 [Accessed 14 11 2020].