Reduction & causation
Causation underlies the workings of the universe and our discourse about it. Anyone who is curious about the natural world must at some time or another in their lives have wondered about the true nature of causation, especially those people with a scientific curiosity. This series of articles on causation became necessary, not only for these reasons, but because causation is so frequently called on to do work in the philosophical debate about reductionism and today’s competing scientific world views.
It is dubious whether the reduction of causal relations to non-causal features has been achieved and scientific accounts are strong alternatives with revisionary non-eliminative accounts finding favour. Can emergent entities play a causa role in the worlds? But is causation confined to the physical realm?
The issue to be addressed here is, firstly, can causation itself be reduced to something simpler. But the role that causation plays in causal interactions that operate within and between domains of knowedge. The outline of this article follows the account given by Humphreys in the Oxford Handbook of Causation of 2009.
At the outset it is important to distinguish between reduction between the objects of investigation themselves (ontological reduction) and linguistic or conceptual reduction as the reduction of our representations of those objects.
Reduction of causation itself
Eliminative reduction of causation
We must decide whether causation is itself amenable to reductive treatment. Reduction may be eliminative reduction in which the reduced entity is considered dispensable because inaccessible (Hume’s claim that we do not experience causal connection) so we can therefore eliminate it from our theoretical discourse and/or real objects (ontology) (the Mill/Ramsay/Lewis model) substituting phenomena that are more amenable to direct empirical inspection. The most popular theory of this kind is Humean lawlike regularity but in this group would be the logical positivists, logical empiricists (e.g. Ernest Nagel, Carl Hempel), Bertrand Russell, and many contemporary physicalists with an empiricist epistemology. Hume’s view was that we arrive at cause through the habit of association and in this way he removed causal necessity from the world by giving it a psychological foundation. A benign expression of this view would be that ‘C caused E when from initial conditions A described using law-like statements it can be deduced that E’.
Non-eliminative reduction of causation
Causation is so central to everyday explanation, scientific experiment, and action that many have adopted a non-eliminative position. X is reduced to Y but not eliminated, simply expressed in different concepts like probabilities, interventions, or lawlike regularities. Non-eliminativists like the late Australian philosopher David Armstrong hold that causation is essentially a primitive concept that we can at least sometimes access epistemically as contingent relations of nomic necessity among universals and thus amenable to multiple realization. language or with eliminativist accounts explaining causation in non-causal terms.
Revisionary reduction of causation
Here the reduced concept is modified somewhat, as when folk causation is replaced by scientific causation. Most philosophical and self-conscious accounts of causation are revisionary to a greater or lesser degree.
Many accounts of causation include reference to causation-like factors as occurs with natural necessity, counterfactual conditionals, and dispositions in what has become known as the modal circle. The fact that no fully satisfactory account of causation can totally eliminate the notion of cause itself is support for a primitivist case.
Domains of reduction
Discussions in both science and philosophy refer to ‘levels’ or ‘scales’ or ‘domains’ of both objects and discourse. So physics is overtopped by progressively more complex or inclusive layers of reality such as chemistry, biochemistry, biology, sociology etc. This hierarchically stratified characterization of reality is discussed elsewhere. Here the task is to examine the way causation might operate within and between these different objects and and domains of discourse.
The attempt at reducoing one domain to another is not a straightforward translation as an account must be given of the different objects, terms, theories, laws, properties and their role in causal processes. The preferred theory of causation (whether, say, a singularist or regularity theory) will be pertinent to what kind of causal reduction may be possible.
Relations between domains
Suppose we are engaged in the reduction of a biological process to one in physics and chemistry, say the reduction of Mendelian genetics to biochemistry, then what kinds of causal interactions might we invoke? The causal relation might be: a relation of identity; an explicit definition; an implicit definition via a theory; a contingent statement of a lawlike connection; a elation of natural or metaphysical necessitation as in supervenience; an explanatory relation; a relation of emergence; a realization relation; a relation of constitution; even causation itself. If indeed the causation were different in different domains then this might render reduction restricted or impossible. Accounts like counterfactual analysis are domain independent.(p. 636)
However, there are domain-specific claims such as physicalism’s Humean supervenience. Under some theories causation is restricted to physical causation as the transfer of conserved physical quantities and this is difficult to apply to the social sciences.
Domain-specific causation & physicalism
Could it be that causation in biology is different from that in physics or sociology or is causation of the same general kind – is their ‘social cause’ and ‘biological cause’ or just ’cause’? The most contentious area here is mental causation where intentionality is often treated as ‘agency’ rather than ‘event’ causation.
In the 1960s domain reduction was promoted through the reduction of theories via bridging laws (Ernest Nagel). One major challenge for such an approach has been multiple realization whereby something like ‘pain’ can be expressed physically in so many ways that this renders its further reduction unlikely although this has been countered by supervenience accounts. For example Humean supervenience regards the world as the spatio-temporal distribution of locaized physical particulars with everything else including laws of nature and causal relations supervening on this.(p. 639) Supervenience is generally regarded a a non-reductive relation.
Multiple realization characterizes properties in terms of their causal roles. Money is causally realized by coins, cheques, promissory note etc. The role of ‘doorstop’ can be functionally and reducibly defined so not all cases of multiple realization are irreducible, irreducibility needs to be taken case by case. For Kim (1997;1999) ‘Functionalization of a property is both necessary and sufficient for reduction …. it explains why reducible properties are predictable and explainable’. Since almost all properties can be functionalized few need to be candidates for emergent properties (p. 644)
Upward & downward causation
The restriction of cause to physical domains is supported by the downward causation and exclusion argument.
Causal exclusion principle & non-reductive physicalism
The causal exclusion principle states that there cannot be more than one sufficient cause for an effect. If we accept this then how are we to account for the causes we allocate at large scales, say the cause of a rise in interest rates? What is the causal relevance of multiply realizable or functional properties (redness, pain, and mental properties)? Does this principle automatically devolve into smallism, that we ultimately explain everything all the way down to leptons and bosons, or smaller and more basic entities when we find them because they are the ones doing the causal work? How can a macro situation have causal relevance if it can be fully accounted for at the micro scale. These properties then become epiphenomena, a by-product or phenomenon with no physical basis.
If C is causally sufficient for E then any other event D is causally irrelevant. Every physical event E has a physical event C causally sufficient for E. If event D supervenes on C then D is distinct from C.
There is increasing evidence supporting the causal autonomy of disciplinary discourse or non-reductive physicalism. Properties in the special sciences are not identical to physical properties since they are multiply realized although they do supervene on (instances of) physical properties since changes in the special properties entail changes in the physical properties further the special properties are causes and effects of other special properties.
A large-scale cause can exclude a small-scale cause. Pain might cause screaming while there is no equivalent neural property. This occurs when the trigger is extrinsic to the system. The pain resulting from a pin prick is initiated by the pin; it cannot possibly be initiated at the neural scale.
The exclusion principle can be applied to any kind of event that supervenes on physical events and shpows that there is no clear causal role for supervening events.
The main questions to be addressed in relation to causation and reduction are: can causation itself be reduced; is there a base-level physicochemical causation underlying all other forms of causation; how does causation operate within a. non-physicochemical domains of discourse and scales and b. between non-physicochemical domains of discourse and scales.
In posing these questions it should be noted that it is cutomary to discuss different academic disciplines, as different domains of knowledge that use their own specific terminology, theories and principles. So for example we have physics, chemistry, biology, and sociology being refereed to as ‘domains of discourse’ and stratified or into ‘levels’ or ‘scales’ of existence. From the outset a careful distinction must be made between ontological reduction, the reductive relations between objects themselves, and linguistic or conceptual reduction which deals with our representations of these objects.
Cause & reductionism
So far in discussing reductionism it has been noted that at present we explain the world scientifically using several scales or perspectives. These scales correspond approximately to particular specialised academic disciplines with their own objects of study including their terminologies, theories, and principles. One possible way of expressing this would be: matter, energy, motion, and force (physics), living organisms (biology), behaviour (psychology), and society (sociology, politics, economics). Each discipline has its own specialist objects of study like be quarks (physics), lungs (biology), desires (psychology), and interest rates (economics). Since it has been argued that each dispiplines is addressing the same physical reality from different perspectives or scales the question arises as to the causal relationships between these various objects of study. This raises the question about the relationship between causes at different scales, perspectives, or, in the old terminology, ‘levels of organisation’ when they deal with different entities. How do we reconcile causation at the fundamental particle scale with causation at the political scale assuming the physical reality that they are dealing with is the same?
To answer this question we need to do some groundwork … our modest philosophical program is to ask: What is causation and in what sense does it exist? Is it something that exists idependently of us and, if not, in what way does in depend on us? Is causation part of the human-centred manifest image? What role does causation play in our reasoning? In other words we need to demonstrate that causation is either a fundamental fact of the universe, or some kind of mental construct, or it can be explained in different and simpler terms.
If we assume the process of explanation proceeding by analysis or synthesis and we regard fermions and bosons as the smallest units of matter then causation must act primarily from the wider context. A rise in interest rates, or the pumping of a heart cannot be initiated by fermions and bosons themselves. To make sense of the fermions and bosons that exist in a heart we must consider their wider context.
Does causation occurs at all scales depending on its initiators or is there a privileged foundational with macroscales explained by microscales, that genes coding (in humans about 25,000 genes and 100,000 proteins) for proteins, cells, tissues, organs, and the organism. That is, a causal chain that leads to progressively larger, more inclusive, and complex structures. This is the central dogma of genetic determinism. But does causation occur between cells, organs, or tissues? Are genes triggered by transcription factors that turn them on and off. Is the environment causal from outside the organism along with other constraining factors at all scales. Homeostasis. evolution occurs through changes in the genotype that are produced by selection of the phenotype as natural selecrtion expresses the organism-environment continuum.
If ‘levels’ or ‘scales’ do not exist as separate physical objects then there is only one fundamental mode of being. This is simply one physical reality that can be interpreted or explained in different ways: it has no foundationalscale or level.
Weak emergence: descriptions at scale X are shorthand for those at scale Y; strong when X cannot be given for Y.
Universal laws apply to biology, an unsupported elephant will fall to the ground, but biology has its own causal regularities that are, of their very nature, restricted to living organisms.
A cause can be sufficient for its effect but not necessary (a piece of glass C starting a fire E) – we can infer E from C but not vice-versa; it may be necessary but not sufficient (presence of oxygen C in a fire-prone region E) – we can infer E from C but not vice-versa. Under this characterization cause can be defined as either sufficient conditions (or even necesary and sufficient conditions).
Some scales of explanation or causal description are more appropriate than others. It is possible to provide an explanation that is either overly general or overly detailed. What is appropriate depends on the causal structure, what would provide the most effective terms and structures for empirical investigation. This contrasts with the view that there is a fundamental or foundational scale at which explanation is most complete. (Woodward 2009). Causes need to be appropriate to their effects. Bosons nfluencing interest rates. Interest rates affecting the configuration of sub-atomic particles. Fine-grained explanations may be more stable but not always. (Woodward 2009).
One are where this tension expresses itself is in the argument over the mechanism of biological selection in evolution. Should we regard natural selection as ultimately and inevitably a consequemce of what is going on in the genes (see Richard Dawkins book The Selfish Gene) or are there causal influences that operate between cells, between tissues, between individuals, between populations, and in relation to causes generated by the environment?
Noble, D. 2012. A Theory of Biological Relativity. Interface Focus 2: 55-64.
It is widely assumed that large-scale causes can be reduced to small-scale causes, the macro to micro: that macro causation frequently (but not always) falls under micro laws of nature. This presupposes a means of correlating the relata at the different scales. This might be interpreted as microdeterminism, th eclaim that th macro world is a consequence of the micro world. The causal order of the macro world emerges out of the causal order of the micro world. A strict interpretation might be that a macro causal relation exists between two events when there are micro descriptions of the events instantiating a physical law of nature and a more relaxed version that there are are causal relations between events that supervene. It might also be the case that even if there is causal sufficiency and completeness the existence of necessitating lawful microdeterminism (laws) does not entail causal completeness. Perhaps in some cases there is counterfactual dependence at the macro but not the micro scale.
Granularity & reductionism
We are tempted to think that we can improve on the precision of causal explanations. Could or should we try to improve the precision of of causal explanations by giving more detail or being more scientific? For example I might explain how driving over a dog was related to my personal psychology, the biochemical activity going on in my brain, the politics of the suburb where the accident occurred and so on. That is, the explanation could be given using language and concepts taken from different domains of knowledge: psychology, politics, sociology, biochemistry and so on. The same situation can be described using different domains of knowledge, scales of existence, and so on. What is of special interest is that the cause will be different depending on the perspective chosen. For simplicity the choice of detail chosen for the explanation is referred to as its granularity. This raises the problems of reduction that is discussed elsewhere. Is there a foundational or more informative scale or terminology that can be used? Is an explanation taken to the smallest possible physical scale the best explanation? Are the causal relations dependant on more metaphysically basic facts like fundamental laws? Do facts about organisms beneficially reduce to biochemical facts … and so on. Is fine grain the best?
Principle 3 – Any description of causation presents the metaphysical challenge of selecting the grain of the terms and conditions to be employed
We can appear to express the same cause using different terms that seem to alter the meaning and therefore the causal relations under consideration, for example: we might replace ‘The match caused the fire’ with ‘Friction acting on phosphorus produced a flame that caused the fire’. This raises the question ‘But what was really the cause?’ with the potential for seemingly different answers when we want only one. The depth of detail in terminology is sometimes referred to as granularity and it raises the question of whether some explanations are more basic or fundamental that others, that some statements can be beneficially reduced to others (reductionism).
This gives us an extended definition of science: science studies the order of the world by investigating causal processes. Causal processes are of many kinds: there are, we might say for example, that Though contentious we might add that we must resist the temptation to reduce causes of one kind to causes of another kind. Causally it makes no sense to reduce biology to physics by saying that fermions and bosons cause the heart to beat. A heart might consist of fermions and bosons but these do not have causal efficacy in this sense. This takes us away from the traditional method of attempting to define science which has been in terms of its methodology (the hypothetico-deductive or deductive-nomological method).
Physicalists can be divided into two camps: those that think everything can be reduced to physics (reductive physicalists) and those that do not (nonreductive physicalists). The reductionist physicalist claims a type-identity thesis such that, for example, mental properties like feelings are identical with physical properties: that all mental properties are caused by physical properties. Assuming we have two entities, one acting causally on the other seems mistaken the two being, in fact, one and the same. Similarly the non causal connection between temperature and mean molecular kinetic energy. Also life and complex biochemistry? The question arises though as to the identity of objects. Is pain physically identical in a human and a herring? Here it seems that pain can be expressed in many different physical ways, known as ‘multiple realization’. This attack on the type-identity thesis led to the modified claim that mental states are identifiable with functional states which then allows multiple realization, a functional property being understood in terms of the causal role it plays. However, we can think of pain as being either coarse-grained, or fine-grained. ??Either one thing, a mix of properties hardly warranting aggregation under a single category, or OK.
Reduction is generally contrasted with emergence. Acounts of emergence are rarely causal in form. Why cannot ‘horizontal’ causation give rise to emergent features within the same domain?