Understanding Compositional Explanations in the Sciences

Understanding the nature of “vertical” relations whether in science, nature, mathematics, logic, or anywhere else, is a hot topic in philosophy. What is unfortunate is that, as yet, too little attention is paid to focused issues about what frameworks work best for the “vertical” relations in particular areas. However, it is very hard to see how one can successfully understand contemporary debates about reduction/emergence in the sciences without understanding the vertical relations posited in compositional explanations. For, as outlined in my last post, compositional explanations are central to such debates — and are plausibly our main (sole?) evidence about verticality in nature.

My concern in the second half of Part I of the book is obviously “focused” in engaging  what I term “scientific composition” relations that are posited in successful compositional explanations across the sciences. Getting an adequate account of scientific composition, and compositional explanation, is crucially important or we will end up distorting the ideas, arguments and positions of scientific reductionism and emergentism, and also mischaracterize the deeper issues. My approach is again to directly engage the explanations and their features, and then to construct an account capturing these notions.

Compositional explanations have variously been termed “functional”, “reductive” and “constitutive mechanistic” explanation. I noted examples in the last post, but the diagram heading the post illustrates a recent case. We explain the process of an ion channel opening in terms of the movements of the protein subunits (show in the diagram) taken to compose these processes. And we explain various properties of ion channels using properties and relations of the protein subunits taken to be their constituents.

Let me start by noting a couple of distinctive, and singular, features, of such explanations.

First, compositional explanations allow us to explain one kind of entity, such as an ion channel or its opening, in terms of the qualitatively different kinds of entity taken to compose it, like molecules or molecular processes. This results in what I term the ‘Piercing Explanatory Power’, or PEP, of compositional explanations. Thus we can explain the entities of the Manifest Image, like the hardness of a diamond, using utterly different kinds of entity, such as the covalent bonding of carbon atoms. But we have now gone on to iterate compositional explanations across all the levels of entities, whether in cytology or chemistry, we have now discovered in addition to those of the Manifest Image.

Second, we should mark that once we have successfully supplied a compositional explanation of certain entities in terms of certain others that compose them, then we have established that these entities are in some sense the same. This is what I will term the ‘Ontologically Unifying Power’, or ‘OUP’, of compositional explanations. For example, at a certain point it was an open question how digestion and molecular processes were related. But once we compositionally explained the process of digestion in terms of molecular processes that compose it, then we established that these processes are in some sense the same. Most importantly, as the latter case illustrates, a successful compositional explanation shows that the mass-energy, or force, associated with a certain higher level entity just is the mass-energy, or force, of certain other lower level component entities. Thus supplying a compositional explanation of digestion, for instance, did away with any need to posit special, uncomposed, Vital energies or forces with digestion.

Compositional explanations, and their distinctive features, are underpinned by relations of scientific composition that back such explanations. Looking at cases, I outline how there are a number of relations of scientific composition holding between a variety of different ontological categories of entity, i.e. “comprising” between powers, “realization” between properties, “constitution” or “part-whole” between individuals, and “implementation” processes. And I highlight fifteen common features distinctive of all of these scientific notions.

This fine grained engagement with actual cases of compositional explanation, and their concepts, is what drives my own account – and my critique of extant rivals. The details of the science matters greatly, but I cannot go through all the points here. Let therefore briefly note a few of the features of scientific composition that are most tightly connected to OUP and PEP, since I use these below to sketch my larger positive and negative points.

Ontologically Unifying Power: I show that the relata of compositional relations are in the some sense the same. I note this is concretized in the sciences in what I term “mass-energy neutrality”. A relation is mass-energy neutral when its relata have mass-energy, but the overall mass-energy of the relata of this relation is equal solely to the sum of the relata on just one side of the relation. Thus, for example, the mass-energy of the ion channel just is the mass-energy of the protein sub-units, and the mass-energy of the kidney just is the combined mass-energy of its nephrons, and other tissue, and their relations. By establishing we have relata that are in some sense the same, and which bear mass-energy neutral relations, compositional explanations thus have Ontologically Unifying Power.

Piercing Explanatory Power: I outline how the relata of scientific composition relation are qualitatively different. But I also detail how scientific composition is a many-one relation – many realizers, or implementing processes, or constituents, on some occasion compose one property instance or process or individual. (Note that this plausibly shows composition is not a relation of identity without resort to multiple composition – many cannot be identical to one even without multiplicity of composition.) The joint character of scientific composition apparently underlies its PEP-piness, since although each is different from the composed entity the components together result in it.

Ultimately, I argue that we can best understand scientific composition as relations of what I dub ‘joint role-filling’. Put roughly, I highlight how we have compositional relations in the sciences when we have collectives of entities with certain roles whose relations to each other allow them to jointly fill the qualitatively distinct role individuative of the relevant composed entity. Thus the movements of the subunits fill the role of the ion channels opening. Or the roles of the nephrons fill the role of the kidney. I offer a starting theoretical framework capturing the general nature of joint role filling and I also consequently provide detailed frameworks for the specific relations of parthood between individuals and realization between property/relation instances.

Crucially, the joint role-filling account accommodates both the Piercing Explanatory Power, and Ontologically Unifying Power, of compositional explanation because it accommodates the features of the underlying relations of scientific composition.

Note that when we have entities that fill the role of some other entity, then these entities are in some sense the same. More importantly, a joint role filling relation is plausibly a mass-energy neutral relation – the mass-energy of the entity whose role is filled just is the combined mass-energy of the role fillers. So, joint role filling plausibly accommodates OUP. In addition, joint role-filling is a many-one relation and it is the case that none of the role fillers is qualitatively the same, or similar to, the entity whose role is filled. None of the entities filling the role actually itself plays this role – rather the entities only jointly fill the role. So joint role filling can have qualitatively distinct relata – hence accommodating the PEP-py nature of compositional explanation. We explain the hardness of diamonds, or their scratching, using entities none of which is hard, or scratches, in carbon atoms and the breaking of molecular bonds.

The joint role filling approach thus accommodates both OUP and PEP, but I argue no other account succeeds in this task. Let me rub these points in a little, since a number of readers will wonder why I bother with a new theoretical framework rather than the existing ones they favor. And this point highlights the dangers of using other frameworks to understand reduction/emergence etc.

Looking at the standard ‘functionalist’ accounts that Kim and others use in recent pioneering work on “reduction” we find role-based views similar to the joint role filling view. But this is a distinct role-based view that uses what I term “role playing” That is, we have a one-one relation between qualitatively similar entities where some property instance plays the very role of the realized property instance because the two instance overlap in their powers. This is what is termed the ‘Flat’ or ‘Subset’ view of realization.

My focus has always been on whether the Flat/Subset view of realization, and other accounts, provides a good account of the compositional relations between properties/relations that we find in compositional explanations in science. In earlier work (Gillett (2002), (2010)), I noted that the Flat/Subset view is not many-one, does not have qualitatively different relata and does not relate property instances of different individuals – hence failing to have key features of the relations backing compositional explanations such as those between the properties of ion channel and protein subunits.

However, we can extend this critique, since we now see why the latter points are damaging when we seek to understand the features of the associated explanations. If we take Flat/Subset realization to back compositional explanation, then such explanations must always be what Dennett labeled “homuncular explanation” – that is, an explanation of why this entity has power X using some other entity that has power X. For role playing requires overlap in powers — so a power is always explained by some other entity that has the same power. But such an explanation obviously lacks the distinctive Piercing Explanatory Power of compositional explanation which allows us to explain why one entity has power X using utterly different powers. So we see why the PEP-less Subset/Flat account plausibly fails as an account of scientific composition and the compositional explanations it backs — and dangerously distorts the issues when applied to scientific phenomena arising from compositional explanation.

Let us turn to work in the philosophy of science. A widespread movement in philosophy of science right  now, following Wesley Salmon, is to treat all explanations of singular facts and/or events as backed by causation or causation-like relations. These “neo-Causal” approaches thus take compositional explanation to be backed by causation-like relations. Carl Craver (2007) offers the most developed version of such a position by applying interventionist frameworks to scientific composition claiming that mutual manipulability suffices for the implementation (i.e. compositon) of processes.

In response, in much more detail than these brief comments, I show that Craver’s elegant manipulability account wrongly counts certain Vitalist accounts as involving compositon when they merely involve manipulability of brutely connected, but uncomposed, Vital entities and certain molecules. So this neo-Causal account conflicts with successful scientific practice in its ascriptions of scientific composition. And I show why. Manipulability does not suffice for relata that are in some sense the same or which are in mass-energy neutral relations. So the manipulability account fails to accommodate key features of scientific composition. Consequently, I note that an explanation backed by manipulability relations does not establish that its relata are in some sense the same, nor does it establish a mass-energy neutral relation – hence the manipulability account fails to accommodate the Ontologically Unifying Power of compositional explanation — hence providing an inadequate account of such explanations. (I outline still more foundational problems for manipulability accounts of composition in forthcoming papers, but the latter issues suffice should here).

Versions of the latter points also apply to all the extant neo-Causal treatments of scientific composition and compositional explanation. (They also apply to many “grounding” based views). So I suggest that we have strong reasons to reject both the manipulability account of scientific composition and other neo-Causal views of scientific composition in favor of my joint role filling view. Notice also that these points plausibly show compositional explanation is not causal explanation – and that the joint role filling framework is our best account of such singular non-causal explanations and relations.

Scientific composition, and compositional or “constitutive mechanistic” explanation, are the not the primary focus of this book – I construct a beginning framework for scientific composition so I can reconstruct scientific reductionism and emergentism, as well the ongoing debates in the sciences. I move onto to that work in Parts II and III of the book that I outline in the next two posts.

However, if you are interested in these topics, verticality is “hot” after all!, I do have something else to offer. To plug another book, Ken Aizawa and I have a forthcoming anthology that highlights the pressing need to engage the nature of scientific composition and the variety of different philosophical approaches to verticality: https://www.amazon.com/Scientific-Composition-Metaphysical-Directions-Philosophy/dp/1137562153.

The anthology has a range of interesting, and diverging, takes on “verticality” and scientific composition  from writers in different research traditions in metaphysics, philosophy of science and metaphysics of mind. The Introduction to the book surveys recent work on “verticality” in different areas of philosophy. And my Chapter in that book goes into more detail about the problems of the two noted approaches, but also the related problems with so-called “grounding” accounts of scientific composition as well. (Anyone interested in that Chapter, or the Introduction, should email me. The anthology itself should be out in Fall some time.)