On Haslanger’s Structural Explanation (Draft)

In her 2016 paper, Haslanger proposed an account of structural explanation. Despite its ongoing influence, I find it quite disappointing. The arguments are quite scattered and disorganised, but I’ll try to do justice to her work. I think many of the lines can be made clearer by connecting to debates in the philosophy of science, and for that precise reason, the theory strikes me as very underdeveloped and unconvincing. It is also worth noting that she has twisted some of others’ works.

1. Why Structures?

This might be the part I have the most problems with. To justify the employment of structures in explaining inequality, Haslanger draws multiple analogies from other disciplines. The opening of the paper uses the example of a treat inserted in a ball. Why is it the case that the treat ends up in the gully when we toss the ball? One way to explain is to give a detailed mechanical analysis of the treat alone. A better way to explain, as Haslanger argues, is to regard the treat and the ball as part of the structure, which was thrown and rolled down the hill into the gully.

It is questionable whether physical structure should be treated on a par with social structures. I think extra caution is required when moving across disciplines, as the meaning of “structure” may shift according to context, but generally speaking, the emphasis on the part-whole relationship and the deemphasis of nodes are the commonalities they should all share. In that regard, I don’t think it is problematic for her to talk about structures per se, but one should be mindful of how they are derived.

The physical structure in Haslanger’s example seems to be a bona fide paradigm which causes its parts — the ball and the treat, or whatever objects playing those roles — to behave accordingly, but this is a misconception. The system holds precisely because of certain preconditions, i.e., rigidity, friction, pre-established laws of nature, etc., and were the treat ever too slippery, the structure collapses as the treat would escape the ball during the toss. The fact is that structures do not exist out there — they are often what we have generalised from empirical patterns, with certain conditions to be met.¹

I think this can be better demonstrated by Galileo’s classic experiment. Picture yourself with two balls of the same size, one weighing 1kg and the other weighing 10kg. Which will land faster when both are dropped from the same height? Anyone with some basic physics education will answer that both will land at the same time, but this was not an obvious fact for people at that time. Now we put the two balls tightly together and drop them from a certain height. Should we treat the two as a system? The answer would be yes, as we know they have the same acceleration rates and thus wouldn’t fall apart. But this was also not obvious for people in the past, as they would believe the heavier ball would fall faster. The point is that physical structures are not preexisting entities but rather contingent on laws of nature; for that reason, their explanatory powers are constrained as they are also derivative from those of higher laws of nature. Given that social reality is a lot messier than the scientific one, it is even more dubious to rely on social structures. I will further explore this in §3 and §4.

Another analogy was drawn from Shapiro’s (1997) ante rem mathematical structuralism. Of course, Haslanger has not explicitly referenced any mathematical structures, but she goes on to extensively cite Shapiro, who will consequently use the distinction he made to illustrate his ante rem structuralism. At the core of the distinction is the places-are-offices perspective — nodes in the structure are positions which objects occupy — and the places-are-objects perspective — nodes in the structure have a distinct ontological status. For ante rem structuralism, the distinction is merely perspectival, in the sense that positions may be described as “offices” (roles within an abstract structure) or as “objects” of an abstract ontological kind. But this shift does not introduce a new metaphysics; both perspectives describe the same structural ontology. Both von Neumann (0 = ∅, 1 = {∅}, 2 = {∅, {∅}}, 3 = {∅, {∅}, {∅, {∅}}}, …) and Zermelo (0 = ∅, 1 = {∅}, 2 = {{∅}}, 3 = {{{∅}}}, …) systems instantiate the natural number structure <N, 0, s>, and the natural number structure trivially instantiates itself as well. The upshot is that background ontology need not be primitive objects.

It is thus puzzling why Haslanger invoked Shapiro in the first place, as the original point is merely to clarify the conception of ontology in ante rem structuralism. Her own employment of the distinction is that the places-are-offices perspective refers to occupiers of positions, and the places-are-objects perspective emphasises the relationships between places. She then uses this distinction to argue that the latter serves as a better explanation than the former. This is a terrible misinterpretation of Shapiro’s texts, which is not intended as an explanatory divide and does not attribute substantive properties to positions. And for that reason, I don’t see how this will lend any support to her invoking social structures in explanations.

I think one of the relatively good examples Haslanger has provided is Garfinkel’s (1981) grading structure. In a class where the instructor stipulates that only one student can get an A, it is not sufficient to answer “Mary did well in the final” to the question “Why did Mary get an A?”; we should instead answer “Mary did the best in the final” to reflect the structure contributing to the grading. The reason why the grading structure is invoked is obvious: the instructor explicitly dictated so, and unless there are, for instance, two equally compelling works which made the instructor decide to break the original rule, the structure can serve as a good source of explanation. However, do all explanations of social phenomena follow a similar pattern where structures are clearly dictated? I’ll leave that question to §4. Some further considerations of explanation will be explored in §2 and §3. But at this stage, I think we can conclude that Haslanger is not justified in using structures from other disciplines to support her own, and the overall worry can be summarised as: what grounds the existence of social structures when we need them in explanation?

2. Triggering and Structuring Causes

This may be the most confusing part of the paper. Before I embark on my arguments, I have to flag Haslanger’s assumption of an erotetic model of explanation, where explanations are answers to questions, which I will also take for granted. This thus lays the ground for the consequent discussion of triggering and structuring causes. The distinction was originally made by Dretske (1988, 2010), where he merely intended it to explain behaviours in philosophy of mind and action, but Haslanger extends it further to the purview of social explanation.

Before I ever diagnose Haslanger’s own application of Dretske’s model, I must make some clarifications and challenges to this distinction. It is worth pointing out that Dretske’s later writing on the distinction diverges from his early writing. In his 1988 work, he said:

In looking for the cause of a process, we are sometimes looking for the triggering event: what causes the C which caused the M. At other times we are looking for the event or events that shaped or structured the process: what caused C to cause M rather than something else. The first type of cause, the triggering cause, causes the process to occur now. The second type of cause, the structuring cause, is responsible for its being this process, one having M as its product, that occurs now… There is a clear difference between explaining why, on the one hand, Clyde stood up then and explaining, on the other hand, why what he did then was stand up (why he stood up then). He stood up then because that was when the queen entered, or when he saw the queen enter, the room. He stood up then as a gesture of respect. The difference between citing the triggering cause of a process (the cause of the C which causes M) and what I have been calling its structuring cause (the cause of C’s causing M) reflects this difference… (Dretske 1988, pp. 42–43)

This is what Haslanger (2016) cited in her own paper. We can roughly picture the model as follows. For an internal event C and its product event M, B is the behaviour C → M. So,

[if] we think of the process in question as C’s causing M, the cause, in this second way of thinking about the cause, is whatever event causes C. So, for example, if a stimulus S produces C, and if C, in turn, causes M, then S, by triggering the sequence, causes the process to occur. (ibid., p.42)

Therefore, the triggering cause is the immediate stimulus of the internal event, i.e., S as in S → C → M, which causes B to occur now. Dretske’s framing of structuring causes, on the other hand, carries a contrastive nature in his 1988 writing, as seen in the first quote above (“what caused C to cause M rather than something else”). The Clyde example demonstrates this: what causes Clyde to stand up when the Queen entered (rather than bowing, kneeling, etc.)? To answer this, we must invoke the structural cause, namely Clyde’s respect for the Queen, as well as the social norms which require him to act in such a way.

I will later illustrate that Haslanger’s argument relies heavily on this contrastive nature of structuring causes, but for now, we shall restrict our focus to Dretske’s later writing. Whilst Dretske’s 2010 framing of triggering causes remains the same, he ceases to talk about the contrastive nature of structuring causes and shifts towards its background and historical explanatory role. There, he gives an example of the terrorist attack: the terrorist plants a bomb in the general’s car, and the general turns the key, which detonates the bomb. He then writes:

When C is the triggering cause of E, one can think of the structuring cause of E as a (triggering) cause of those standing conditions (call them B) in which C causes E. The terrorist’s activities a week ago, A, produced (or are the triggering cause of) those conditions of the automobile, B (bomb wired to the ignition), in which turning the key in the ignition (C) causes an explosion (E). So the triggering cause of B is a structuring cause of E. (Dretske, 2010, p.140)

Dretske’s point can be put more clearly this way: the general’s turning of the key is the triggering cause of his death, but the structuring cause is the terrorist’s earlier action. By wiring the bomb into the ignition a week before, the terrorist created the standing conditions in which the key-turn would produce an explosion. The key-turn causes the explosion only because those background conditions were already in place. Thus, setting up the standing conditions, i.e., the historical and physical arrangement that makes a certain outcome possible, is what Dretske means by a structuring cause.

Where did the contrastive feature go? Under an erotetic model of explanation, there is a strong appeal in contrastive questions, where asking why E rather than E′ often yields more illumination than simply asking why E, plus, as Lipton (1990) argues, most why‑questions are implicitly contrastive. But even on a contrastive view, what ultimately does the explanatory work are the underlying causal histories and standing conditions that make E occur instead of its alternatives. Contrastive framing serves to pick out which portions of that history are relevant to explaining a given “Why E rather than E′?” question. It is indispensable at the level of explanation‑as‑answering‑a‑question, but it does not introduce a new kind of causal relation over and above the ordinary causal set‑up. Read this way, Dretske’s later talk of structuring cause is best construed as highlighting the historical and background conditions that fix which outcome a given triggering cause produces, with the contrastive element now embedded in the specification of those conditions rather than explicitly spelt out. Therefore, in the terrorist’s case, although we may raise the contrastive question “why turning the key causes the car to explode rather than mere ignition”, it should be borne in mind that the contrastive form stems from a deeper desire to understand the causal history more completely, which a simple “key-turn” answer could not possibly capture.

Now I wish to move on to what Haslanger makes of Dretske’s texts. After citing Dretske’s text on the Clyde example, she goes on to say:

Presumably this would be to situate Clyde within a social structure: Clyde is Queen Elizabeth’s subject (he is a part in a larger whole). Locating Clyde in a structure, we shift our attention from individuals to nodes: subjects are required as a matter of etiquette (the structural conditions that limit the possibilities for nodes in the structure) to stand when the Queen enters. This is not a quirk of Clyde’s: generally subjects of the Queen will stand up when she enters the room, given the relations between Queen, her subject and the rules of etiquette. Once we make explicit the background structure that constrains a subject’s (in this case, Clyde’s) behavior, we can then ask, why is standing up the polite thing to do? And why are there Queens and subjects at all? (Haslanger, 2016, p.121)

I wouldn’t doubt that if this is only intended to further Dretske’s point, i.e., more is needed to account for Clyde’s behaviour than the simple “the Queen enters” answer, then nothing essentially goes wrong; this says nothing profound either. However, Haslanger is clearly aiming at something else.

3. The Nature of Explanation

How can a structure explain? In philosophy of science, and maybe also philosophy of social science, many models of explanation have been proposed in the past — Deductive-Nomological Model, Statistical Relevance Model, Causal Model, and so on. It is clear that Haslanger is using the causal model here.

4. Structural Explanation in Practice

Haslanger’s most famous example of structural explanation might be the “invisible foot”.

Footnotes