Redescription,Reduction,and Emergence

I would like to thank Tobias Hansson Wahlberg for his careful attention to my work in his recent paper in this journal (2013), which focuses on some central elements of my argument for the existence and causal powers of social entities or objects (Elder-Vass 2010). However, I wish to correct some apparent misunderstandings of my argument and its significance.

As Hansson Wahlberg makes clear, I claim that social entities or objects have emergent causal powers in what is often considered to be a weak sense of emergence (e.g., Stephan 2002, 79): that they have powers their parts would not have, even collectively, if they were not organized into the relevant sort of whole by virtue of a characteristic set of relations that obtains when parts of this sort are organized into a whole of this sort. I offer justifications for this as a general ontological perspective and seek to apply it to social entities composed (at least in part) of individual human beings, including, for example, organizations, queues, and norm circles. Although he does discuss some of these examples, the thrust of his criticism is directed against the more abstract aspects of the argument, and this response focuses on what I take to be the two main criticisms. The first, I will suggest, is simply an error of interpretation, whereas the second is more significant as it reflects the views of a range of other philosophers.

Hansson Wahlberg begins by discussing what I have called the redescription principle. The relevant paragraph from Elder-Vass (2010) states,

This is the principle that if we explain a causal power in terms of (i) the parts of an entity H; plus (ii) the relations between those parts that pertain only when they are organised into the form of an H; then because we have explained the power in terms of a combination—the parts and relations—that exists only when an H exists, we have not eliminated H from our explanation. The entities that are H’s parts would not have this causal power if they were not organised into an H, hence it is a causal power of H and not of the parts. The lower level account of H’s powers merely redescribes the whole, which remains implicit in the explanation. In other words “upper- and lower-level accounts refer to the same thing, as a whole and as a set of configured interacting parts” (Wimsatt 2006, 450) and hence a casual explanation which invokes the set of configured interacting parts implicitly invokes the same ontological structure as one that invokes the whole. As Geoff Hodgson has put it, in a discussion of methodological individualism, “explanations in terms of individuals plus relations between them amounts to the introduction of social structure alongside individuals in the explanantia.” (Elder-Vass 2010, 24-25; Hodgson 2007, 211)

Hansson Wahlberg doubts that the redescription principle can be sustained on the grounds that identity statements are invalid when they identify singular entities with pluralities. Presumably, this comment is aimed at my use (and indeed Wimsatt’s) of “the same” in the quoted paragraph (I have also used similar locutions elsewhere) and the identification of the parts (and connecting relations) of a whole as a plurality. After several pages of criticism based on this claim, he suggests that if I wish “to avoid eliminating social objects . . . [I] must say that they are either: composed of individuals interrelated in appropriate ways (where the individuals and the relations may differ over time) . . .” [and he offers an or that is not relevant to this response](Hansson Wahlberg 2014, 782). Composition, however, is not “the same thing as identity” (Hansson Wahlberg 2014, 783).

In using the phrase “the same ontological structure,” however, I did not intend to claim that the identity conditions over time of a set of parts are the same as those of a corresponding whole. I hope that I can correct any misunderstandings by making this clear now. My argument is that (a) a set of configured interacting parts that compose a given whole at a given time, including both the parts themselves and the relations between them, is necessarily indistinguishable in terms of its causal capacities from (b) the whole itself at that same time. Thus, an explanation of an apparent causal effect of that whole as an effect of that set of parts and relations is one that fails to eliminate the whole itself from the explanation, since it merely substitutes an alternative description of that whole as it exists at that moment in time for the concept of the whole itself in the explanation. This is thoroughly consistent with the view, which I also hold, that social objects are indeed “composed of individuals interrelated in appropriate ways (where the individuals and the relations may differ over time)” (Hansson Wahlberg 2014, 782), and I discuss the issues involved in changes in objects and social objects at several points in the book, in ways that should make it apparent that I do hold this view (notably Elder-Vass 2010, 35-38 and 133-38). Hansson Wahlberg’s discussion of identity theory is thus irrelevant to my argument (and incidentally also rests on a somewhat perverse reading of it).

As Hansson Wahlberg goes on to say, I argue that if we are to make sense of the causal powers of the parts of a whole, we must consider what causal impacts they could have in a counterfactual scenario “in which the higher-level object no longer exists because the required organization of its parts is lacking” (Hansson Wahlberg 2014, 784). Any further causal powers beyond these, which exist when the higher-level object does but not otherwise, are causal powers, and relationally emergent properties, of the higher-level object in its own right. Hansson Wahlberg, however, dismisses such powers as “diluted and uninteresting” (Hansson Wahlberg 2014, 784). He takes the view that for a social object to have a causal power of its own at time t, this would have to be a power that was “not merely the sum (or net effects) of the impacts of its proper parts interrelated as they in fact are interrelated at t” (Hansson Wahlberg 2014, 783)—what is generally thought of as a strongly emergent property (Stephan 2002, 79).

One of his reasons for taking my view of causal powers to be diluted and uninteresting is derived from the discussion of identity covered above. The second, however, is that “no one denies that the way entities are arranged and interrelated makes a difference to causation, and to suppose otherwise is to turn one’s attention to a straw man” (Hansson Wahlberg 2014, 784). This is closely related to an earlier comment that

no sensible ontological individualist will deny that people can be organized in certain ways, interact, and enter into the kinds of relation with one another with which Elder-Vass is concerned, and thereby cause things they would otherwise not have caused. (Hansson Wahlberg 2014, 778)

In making such claims, he is thoroughly in tune with mainstream thinking in the philosophy of mind, which has frequently seen only strong versions of emergence theory as worth investigating, and with individualists in the social sciences, who think they can make use of organization and relations in their explanations without thereby committing themselves to the belief that social objects have powers. In a particularly perceptive discussion of the same book, Julie Zahle has suggested that this is the most problematic question for my argument: why are individualists not entitled to make use of such elements in their explanations? (Zahle, 2014).

Perhaps the simplest response I can make to that question is to say that if individualists do wish to make use of them, they need to justify why they can. Individualist responses to weak or relational emergentist arguments seem to take for granted that ontological individualism is coherent and justified. Yet they tend to do so without offering any analysis of the nature of causality in general or of how we might justify claims for causal significance. The only argument that seems to be proffered is that social wholes cannot be causally significant because we can offer explanations of their purported causal contributions to social events purely in terms of the causal influence of individuals and the ways in which they are organized and related (a claim that is taken to be true in principle for ontological reasons in every case, even if it happens that in some cases that we do not have a fully adequate explanation in such terms). Yet if this argument can be applied to social wholes, individualists owe us an explanation of why it cannot also be applied to human individuals. If we did so, it would seem to imply that human individuals are not causally significant either because we can in principle explain their actions in terms of the influence of their parts and the ways in which they are organized and related. How, then, can they justify the claim that individuals are real and causally significant while social wholes are not? Unless ontological individualists can offer a coherent answer to this question, their own argument saws off the branch upon which they are sitting (and, of course, because we can apply the argument recursively, it saws off all the lower branches and the tree trunk too, with the consequence that causality disappears from our world altogether, unless there is some bottom level of fundamental particles to which the argument does not apply) (cf. Block 2003).

By contrast, one of the merits of my argument is that it applies equally to both human individuals and social objects. Both, I argue, are wholes with relationally emergent causal powers that differ from the independent causal powers of their parts, and both as a consequence are causally significant. Unlike the arguments for ontological individualism noted earlier, it does seem to provide us with a potentially coherent way of understanding causality in general. Indeed, I would argue that if we wish to make general claims about causality in the social world, then this requires that we have a general theory of causality against which we can assess such claims, ¹ and ontological individualism is incoherent in the absence of such a theory.

No doubt the relational version of emergence theory advocated here is less mysterious, less exciting, and less ambitious than the kind that philosophers of mind have sought for so long. But we have no need for these strong emergence theories, and indeed it seems highly questionable whether there are any emergent properties in this strong sense. ² What we do need is ontological theories that are compatible with the successful practices of actual science, and thus with us living (as we do) in the kind of world in which such practices can be successful. ³ In particular, we need theories that are compatible with two key features of that practice. First, a vast range of entities are accepted as having causal significance, and second, this causal significance is taken to be compatible with the production of explanations of how it arises (Gell-Mann 1995, 112). Relational emergence theory meets both of these requirements and provides us with an ontological framework that recognizes the need for sciences of each level of structure: sciences that recognize which macro structures have which kinds of causal influence and also seek to explain how they can have it. This is all we need from the concept of emergence. And this gives us all that we need by way of resisting eliminative reductionisms like the ontological individualism discussed by Hansson Wahlberg: it enables us to justify the assertion that higher-level entities including social objects have causal powers while resisting the antiscientific insistence that such powers are in some sense uncaused or unexplainable. We can often explain the mechanisms through which the interaction of parts configured as they are in a certain kind of whole produces the causal powers of the whole. The fundamental error of eliminative reductionisms like ontological individualism is to believe that the possibility of offering such explanations entails that wholes and their powers are both causally and explanatorily irrelevant.