Causal Explanation - Explanation in Biology

2. Explanation in Biology

2.1. Causal Explanation

Philosophical theories of explanation aim to explicate what properly constitutes an explanation. At the same time, as an implication, they provide normative criteria for when something proposed as an expla-nation succeeds at being one, as well as criteria for comparing differ-ent explanations. The general idea is that explanation refers to a thing, the explanans (e.g. another event), which is in a proper relation to the explanandum (for example, there is a lawful connection between the types of things that explanandum and explanans are tokens of). Not all explanations are causal explanations, unless a causal theory of explana-tion proves to be a universally true theory of explanaexplana-tion, but for the

purpose of this dissertation, I discuss only causal explanation.²¹ There are, however, two very different approaches to what the aims of a the-ory of explanation should be. I will call them fundamentalist theories and pragmatic theories. In the context of causal explanation, they also imply two quite different approaches to causal relations that can be characterized, likewise, as fundamentalist and pragmatic theories of causation. I will make some remarks on this now, before articulating the main ideas of the contrastive-counterfactual theory of explanation and the accompanying manipulation theory of causality.

2.1.1. The Aims of the Theory of Explanation

What I call the fundamentalist theories are those that aim to explicate how the explanandum fits to the fundamental structure of the world.

According to some such views, the nomological theories, explanation in-volves general laws that describe or reflect the real regularities of the fundamental structure or processes in the world and from which the explanandum can be inferred (Hempel 1965; Salmon 1971; Friedman 1974; Kitcher 1989). According to others – call them causalists – expla-nations refer to ontologically fundamental causal powers and capaci-ties, from which the explanans selects those that are relevant (Mackie 1974; Harré & Madden 1975; Salmon 1984; Cartwright 1994). Accord-ing to these fundamentalist views, an individual object of explanation is fully explained when we know exactly how it fits to the way the world works in general. Discovering the fundamental

change-21 A causal theory of explanation (e.g. Salmon 1971; Salmon 1984) makes the claim that a thing is explained when we know what caused it. This reduces all forms of explanation to causation, and by implication theories of this sort are theories of causation first and theories of explanation second. A theory of causal explanation (e.g. Ylikoski 2001; Woodward 2003), in contrast, is a theory that explicates a causal explanation without needing to be a theory of causation, to presuppose any specific ontology for causation, or making a generalization about all explanation being causal.

producing structure of the world (whether it includes only physical laws or also some emergent causal powers) is an important aim of sci-ence – whether we can do so or not. It is also an interesting philosoph-ical question to ask what fitting an individual event into a fundamen-tal structure like this would consist of. This cannot be a general theory of explanation, however, for the following reasons.

First, a theory that aims to explicate what takes place at the fun-damental level of connectedness²² between the things in the world is mostly unusable. It cannot be used to describe and normatively eval-uate most of the actual practices of scientific explanation when, for example, we are interested in finding the right kind of systematic de-pendencies between two things of interest on a domain of research that is not at the level of these fundamental dependencies. In other

22 There are multiple ways of understanding “levels”, as discussed in the Intro-duction, and this applies here, too. One way to understand “more fundamental level” is to think about it in terms of ontological levels: some things function as the ontological constituent parts or “grounders” of the higher levels. There is a persistent intuition that more fundamental and more constitutive entities are also smaller: the entities of lower ontological levels are also the structural component parts of the systems of higher ontological levels. The idea of a mechanistic under-standing of the ontology of levels (see Wimsatt 2007; Craver 2007; Levy 2013) sometimes rests on this intuition, but they are at least conceptually separate (Craver 2007 & 2015; Kuorikoski 2009). “Levels of explanation” can refer to either the hierarchical levels of constitution (like in the case of individuals and groups in the multilevel selection models) but also to the abstract levels of mechanism: the constituent parts of a mechanism need not be on the same ontological or hierar-chical level with each other (Craver 2007). If this is adopted, what is “more funda-mental” in mechanistic explanation is not the same as “more fundafunda-mental” in a structural or ontological analysis. Furthermore, levels may be levels of abstraction for analytical purposes even in explanation (like the functional and computational levels in cognitive science; see Marr 1982). The traditional way to think about fun-damentality of levels in philosophy posits the funfun-damentality of “fundamental physics”, which is usually thought to be fundamental in all the relevant senses.

For the current discussion, I grant the existence of such a fundamental level, what-ever this means. If there is no such thing, fundamentalist theories automatically fail.

words, it is not a suitable approach to explanation in special sciences, including biology. We need a philosophical theory for characteriza-tion and for criteria of evaluacharacteriza-tion here too. Second, things are con-nected in complicated systems in complicated ways, and looking for the fundamental connectedness (such as physical causal processes; for example, Salomon 1984) is not necessary, may even be misleading (see Woodward 2003), and does not reflect the explanatory practices in special sciences. When we deal with “imperfect” or “elliptical” expla-nations that do not give the whole picture, we need philosophical tools to compare different explanations, even if they are all true in their domain, to make sense of the whole picture and to relate the var-ious explanations to each other. For example, we need tools to map the differences in what explanatory facts they point to and what the relation between these facts is. This calls for pragmatic theories of ex-planation – either to replace or, at least, to complement the fundamen-talist theories.

The pragmatic approach to the theory of explanation takes the practice of explanation as its starting point for explication, clarification, and, subsequently, sophistication. It aims at a normative theory of a successful explanation without making a reference to a fundamental connectedness as a necessary criterion. This approach does not con-tradict the principal idea of fundamentalist theories that the real de-pendencies that the causal explanations trace are constituted by the fundamental structure of the world and how the change in it works – they are mostly neutral about this. It may be that some of the funda-mental theories are compatible with some of the pragmatic theories, giving explications for different notions of explanation, instead of be-ing rivals. One could, for example, consider pragmatic theories to be the theories of “partial” explanations, even if one thinks that a fundamen-tal theory is needed to account for deeper questions of what ultimately makes things explanatory. The pragmatic approaches take the role of explanation, and therefore the role of the theory of explanation, to be more modest and practical than tracing fundamental structures exclu-sively. The notion of explanation in this dissertation is a pragmatic

one. I will not argue against fundamental theories, but they are irrele-vant to my discussion.

The old school pragmatic theories were theories of the pragmatics of explanation (Gärdenfors 1980; van Fraassen 1980; Tuomela 1980;

Achinstein 1983; Sintonen 1984). They aimed to explicate the explana-tory questions and give criteria for practical evaluation of whether the given explanatory information was an answer to the explanatory question or not. Newer pragmatic theories, in contrast, aim at a nor-mative theory of what counts as explanatory information and for what.

The philosophical work in the newer theories is strongly based on the work done in the earlier theories, but the aims are different. Conse-quently, the new pragmatic theories include criteria for what kind of things in the world are explanatory – even if they are not theories of what make things ultimately explanatory on the fundamental level.

There are two approaches to this. One is the contrastive-counterfactual theory (Woodward 2000 & 2003a; Woodward & Hitchcock 2003; Yli-koski 2001; YliYli-koski & KuoriYli-koski 2010), which gives an account of how a partial explanation works, based on explanatory interest rela-tive selection of real dependencies. The other is the mechanistic theory (Bechtel & Richardson 1993; Glennan 1996 & 2002b; Machamer, Darden & Craver 2000; Craver 2007), which aims to explicate the logic of explanation as a practice of revealing a type of dependency, a mech-anistic connection that can be used in explanation.²³

The contrastive-counterfactual theory and the mechanistic theory are not exclusive alternatives to each other but approaches concentrating

23 Including mechanistic theories in this category might seem a controversial move: one could take the stance that mechanistic theory is a causalist theory of explanation, if one takes it to be a theory of causation, too. I will discuss this matter shortly. But the main reason for including it here is that the proponents of this theory are interested in the structural connections of mechanistic inter-actions in picking out the explanatory relations between things, instead of the causal powers of the entities that constitute these structures. Furthermore, the mechanistic ideas are used in this thesis as pragmatic, not ontological, ideas about explanation and causation.

on slightly different issues of causation and explanation, and they are often combined (for example, Craver 2007). This is a widely although not universally held position, and it is adopted in this dissertation. I take the contrastive theory to be the primary theory explicating the concept of explanation. The accompanying manipulation account of cau-sation is the primary perspective on causal relations when it comes to the practices of causal explanation in science. The mechanistic theory is an important addition to account for relevant explanatory infor-mation and to speak about the nature of causal relations in biology (see Ylikoski 2001; Woodward 2002 & 2011). These approaches are not only mainstream in the philosophy of science in general, they are all but dominating views within the philosophy of biology, so I will not advance a systematic defence for them. It is, however, worth articulat-ing what these theories say about explanation and some of the reasons why these are the adequate approaches to assume at this point, since these ideas inform the substantial discussion of the main topics.

2.1.2. The Contrastive-Counterfactual Theory of Explanation The starting point of the contrastive-counterfactual theory of causal explanation²⁴ is that science traces real patterns of causal dependencies.

The dependencies do not require a specific direct connection or pro-ductive relation between the cause and the effect. This idea does not contradict the idea that causal relations are ultimately constituted by the fundamental structure of the world and whatever connects events or things to each other spatially. The regularities that are supposedly described by the strict laws of future fundamental physics may pro-vide the description on this level. For any object of study within spe-cial sciences, however, there is a multitude of causal connections in a complex network, and for any particular explanatory interest, only some subset of them is relevant. An explanation is an answer to an

24 This brief review of the main points of the theory is based on Ylikoski 2001

& 2007; Woodward 2003a & 2004; and Woodward & Hitchcock 2003.

explanatory question that focuses on some subset and implicitly con-strains what kind of answers are relevant. For example, if we are in-terested in knowing why a window got broken, we might sometimes be interested in the molecular structure of glass, sometimes in the in-tentions of the child who threw the ball at it. A complete causal de-scription would include both (under some dede-scription) and a lot of other things, but that is not what explanations usually aim at. On the contrary, much (or even most) of the causal information is not even explanatory within the chosen explanatory framework. Real explana-tion is always aspectual. The contrastive-counterfactual theory anal-yses the logic of such a partial explanation and gives criteria for explan-atory relevance. According to it, the proper logical form of an explana-tory question is not just “why x happened”, but “why x¹ happened instead of x²…xⁿ”, where x¹…xⁿ are mutually exclusive alternatives.

x²…xⁿ constitute the contrast class for the explanandum x¹ – therefore the explanatory question is contrastive. The contrast class may be implicit in the explanatory question, but it is analysable in the context, if the question is unambiguous.

For example, if we are interested in explaining a particular be-haviour of an animal, say a cat attacking something small that moves, we are not interested in just anything that contributes causally to its behaviour, nor everything that is. We are interested in something that makes the animal behave in this particular way instead of some other particular ways. The contrasts are either real or theoretically possible alternatives (depending on the explanatory interest), and mutually ex-clusive. Knowing all causal dependencies relevant to the behaviour and contrasting it with all logically possible alternatives would max-imize our explanatory understanding of the behaviour, but in prac-tice, the explanatory interests are always narrower. The adequate al-ternatives could include things like the cat not paying any attention to the moving object at all or shying away from it. The explanatory factor we are after is the difference maker between these alternatives: it is the factor that leads into the explanandum instead of anything else in its contrast class. Different explanatory questions have different contrast

classes and call for different difference makers, even if they are all ex-planations for the same thing, such as the same behaviour. This is es-pecially important in evolutionary explanations that explain a trait as being an adaptation for something. First, an adaptation explanation does not, at least in its proper use, simply say that a trait is useful, it says that it has been (1) more useful (2) for some particular purpose (3) than a set of alternative traits that either existed in the earlier history of the species or could have plausibly emerged instead of the actual trait that is being explained. Whatever the trait’s use is, specified this way, it is a difference-maker. Secondly, an adaptation explanation needs not be a claim that this use was the only causally relevant factor in the trait’s evolution – to give an adaptation explanation is to choose one property of the trait (its fitness-increasing use) special interest.

Whether this is appropriate or not depends on the research question.

There is a sharp distinction between using adaptive value as the ex-planatory focus (exex-planatory adaptationism) and assuming that the adaptive value is all we need to know to understand how the trait emerged (empirical adaptationism; see Godfrey-Smith 2001). I will re-turn to this later.

The explanans must be contrastive too. Even if the explanans only mentions the difference maker, it presupposes mutually exclusive al-ternatives to it: the explanans is of a form “because y¹ happened instead of y²…yⁿ”. This is important for several reasons. First, for practical purposes, the description that picks the cause may be ambivalent re-garding the details mentioned in the description. That is, the relevant causal factors of the explanans may not be specified by a mere reference to the explanans. Making the contrast class of the explanans explicit makes the relevant causal factors explicit as well. The property needs not be specified as such, only implied by the contrast class. However, there is a more substantial reason for doing this. In a complex system, several different changes in the causal history of the system could cause a similar change in the explanandum and different explanatory questions ask for difference-makers within different contrast classes.

This must be made explicit when comparing explanations. Two

alternative causal explanations do not necessarily compete even if they point to two different factors that would make a difference in the same way. They may refer to different “locations” in the system that could be altered, and they may even all be necessary parts of the causal history of the system that brings about the explanandum, even with the same contrast class. This is important to note when compar-ing different explanations and explanatory approaches to human be-haviour, for example: many of them may be complementary and true at the same time, despite superficial discrepancies, while all of them are partial and cannot be treated as more than that. I will return to this in later chapters. Adopting the contrastive theory helps to conceive this, and explicating the contrast class of explanans makes it explicit.

Take the explanation of the instinctive aggression of a cat for an example again. This behaviour as such is the complete target of our explanatory interests, and a full explanatory understanding of it would consist of knowing all the factors on which it depends, but any particular explanation points to a particular factor without which the behaviour would not take place. If we could intervene in the system such that the proposed explanatory factor is changed to one of its al-ternatives (without changing anything else), the behaviour would change,²⁵ but we are not interested in any such intervention. At least in principle, several different interventions in environmental factors, as well as in the animal’s psychological (or neural) states, could have the same effects (similar changes in the behaviour), but these factors are not competing explanations for the typical behaviour, since they are searching for difference-makers within different contrast classes.

In other words, they are mutually inclusive explanations and may

25 It does not matter for the semantic analysis of the logic of explanation whether we could actually make these interventions – we are using the notion of ideal intervention in order to articulate the logic. The actual feasibility of the interventions matters only for whether we can discover the causal dependen-cies through actual experimental settings. The discussion for now is about the semantics of what is means to explain. I will return to the causality part of causal explanation shortly.

point to different factors in the network of causal connections that re-sult in the behaviour. In addition to a direct intervention in the mech-anisms triggering the behaviour, the cat’s “complete” behavioural dis-position could be affected through intervention in its development, which would not only be an intervention in a different factor, but in a different kind of factor. I will return to the various kinds of biological explanations later in this chapter. Again, the relevant change to the psychological (or neural) disposition could take place both internally and externally: by there being different genes or epigenetic differences in developmental pathways, and by the cat growing up in a different environment.²⁶ Another different dimension of causal explanation (with a different difference maker and contrast class) is called for if we are interested in the environmental factors in the evolutionary past

In document Evolving in Groups : Individualism and Holism in Evolutionary Explanation of Human Social Behaviour (sivua 50-89)