No top-down, or bottom-up, causality

I am preparing a paper for what looks like a very promising workshop to be held in May at the Konrad Lorenz Institute for Theoretical Biology near Vienna, on the topic of causes and processes in biology. The workshop is organized by Kevin Laland (St. Andrews, UK) and Tobias Uller (Lund, Sweden), and aims at initiating close interaction and exchange between philosophers of science and biologists to reflect on the nature of causation in biological evolution. The Extended Evolutionary Synthesis — a project in which I am involved — has a different perspective on causation in evolution, and ascribes evolutionary significance to a greater range of processes, than traditional perspectives. As the organizers put it, “the workshop will set out to scrutinize these claims, with both philosophers (acting as independent arbiters) and non-project members (including non-sympathizers) present to ensure good debate.”

My talk will be on “Causality and the role of philosophy of science,” and while preparing the accompanying paper I ran into a very interesting contribution to the debate on causality by Carl Craver and William Bechtel, entitled “Top-down causation without top-down causes.” I think readers of this blog will be interested, so I’m going to present a summary of the Craver and Bechtel paper, with a little commentary to draw out its more general lessons.

Biologists and other scientists often talk about “top-down” causation, apparently without concern for the fact that physicists don’t seem to recognize any such category of causal interactions, instead seeing everything from a bottom-up perspective (which, as we shall see, turns out to be equally problematic!).

The topic often comes up in discussions about free will and determinism, even though as I’ve argued elsewhere, determinism is entirely irrelevant to the issue of free will. Besides, I tend to be agnostic concerning the question of whether we live in a deterministic universe or not, since that depends on how one thinks of the implications of quantum mechanics for the question (my understanding — and I’m not a physicist — is that q.m. equations are deterministic, and yet q.m. phenomenology appears to be irreducibly stochastic). I do, however, subscribe to the existence of universal cause and effect (meaning that every phenomenon has causes), so causality is an important topic for me.

Philosophers are not necessarily wedded to one view or the other about the existence of top-down causation, but are obviously very much interested in the nature and manifestations of causality. They are also typically cautious not to unnecessarily multiply entities or levels of description, attempting to keep their ontology as sparse as it is possible given the actual structure of the world, as W.V.O. Quine maintained we should.

Craver and Bechtel provide a splendid example of rigorous philosophical analysis of the issue, arriving at a sensible position that they, somewhat paradoxically, term top-down causation without top-down causes, whence the title of their paper. These authors’ concern is that when scientists invoke top-down (or bottom-up) causation they call upon a concept that is either incoherent or implies the existence of “spooky” forces that wholes somehow exert over their components (or vice versa).

Their solution to the problem is to grant that talk of top-down or bottom-up causation is perfectly legitimate as a shortcut, as it captures a number of observable relationships between the activities of wholes and the behaviors of (some of) their parts, but they recast the explanation of what is going on in such cases as not actually consisting (entirely) of causal relations. Rather, Craver and Bechtel propose to think in terms of mechanistically mediated effects that are themselves hybrids of two components: i) intra-level causal relationships, and ii) inter-level constitutive relationships.

[Before you ask, yes, they discuss in detail what they mean by “level,” limiting themselves to levels of mechanisms acting in any particular case. They also provide a discussion of different accounts of causality to argue that only intra-level causes are not “spooky.” It will be clearer what they mean when we get through a couple of examples, but the full paper is linked to above, for the reader who wishes to dig deeper.]

Hal the tennis player and top-down causation

Let us consider just one of the several examples of top-down causation presented by Craver and Bechtel in the course of their thorough analysis of the issue. They introduce us to Hal, a tennis player, who steps on the court and begins to serve. While this is happening, at the physiological level the membranes of Hal’s cells absorb the glucose that was at that moment circulating in his blood; once transported inside the cell, the glucose is phosphorylated and bound into molecules of hexosediphosphate.

A very reasonable description of what’s going on — and one that biologists routinely use — is that the fact that Hal, the organism, began to play tennis initiated a cascade of lower level physiological and molecular effects. These processes started because Hal began to play, and so the scenario is a classical example of top-down causation: the behavior of the whole (Hal) has affected the behavior of the parts (cells and metabolites).

What Craver and Bechtel argue, however, is that a more reasonable account of what is going on can rely on a combination of intra-level causation and constitutive effects, doing away with top-down causality altogether. Here is their reconstruction of Hal’s case:

“When Hal started to play tennis, the nerve signals to the muscles caused them to metabolize the available ATP to ADP to provide the energy to contract the muscle cells. The increase in ADP made it available as a receptor for phosphates in high-energy bonds in 1,3-diphosphoglycerate produced at the end of the glycolytic process. This allowed a cascade of reactions earlier in the pathway to proceed, eventually allowing a glucose molecule to take up a phosphate molecule. In this and many similar cases, a change in the activity of the mechanism as a whole just is a change in one or more components of the mechanism which then, through ordinary intra-level causation, causes changes in other components of the mechanism. Hal’s playing tennis is in part constituted by activities at neuromuscular junctions, and activities at those junctions cause, in a perfectly straightforward etiological sense, changes in the organization and behavior of cellular mechanisms. … Once we have described the mechanism mediating the effect, the drive to speak of this as a case of top-down causation vanishes, although such language might be useful as shorthand.”

The heart attack killed the general: bottom-up causation

What about bottom-up causation, so dear to reductionist physicists? Well, that one too is in trouble, for the same reason, i.e., the postulation of inter-level “spooky” causality. Here is one of the examples analyzed by Craver and Bechtel.

Consider a general who is killed by a massive heart attack. Thinking of the general not as someone performing a particular social role in the military, but as a biological being, then it is obvious that the heart is one component, among many, of one of his bodily mechanisms, the circulatory system. A natural way to talk is in terms of bottom-up causation: the failure of the heart causes the failure of the circulatory mechanism, which in turn causes the death of the general. Craver and Bechtel explain:

“By combining an intra-level conception of cause and a constitutive relation between levels, we can provide an unproblematic account of what transpired: we trace the effects of the infarction through the mechanism. When the heart stops beating, it stops transporting oxygen and nutrients to the other tissues of the body, and they cease to function. We thus explain how a variety of physiological mechanisms cease to function. And insofar as that non-functioning constitutes the general’s death, we explain his death. Notice that when we reach the state of the mechanism that constitutes the state of death, we do not say, with Betty Crocker, that it causes death. It just is death. An assassin who charges extra for the general’s death — after charging for the cessation of physiological function — is pulling a fast one. Our hybrid explication thus avoids the temptation to attribute causation between a constituent and the mechanism of which it is part.”

The broader point

Craver and Bechtel discuss several other examples like the ones above, all leading to the same conclusion: any alleged case of top-down or bottom-up causation always resolves into the sum total of two components, intra-level causation plus constitutive effects.

In my mind, Craver and Bechtel’s analysis represents a very good example of one of a series of modes in which philosophy of science aids science (most of it doesn’t, focusing on its broader business of understanding how science works or fails to do so).

In another case I discuss in my contribution to the KLI workshop, concerning the issue of proximate vs ultimate causes in biology (I wrote about that on the blog, here), there was a genuine scientific debate, involving different theoretical positions staked by scientists themselves. Some biologists think that only ultimate causes (natural selection) are necessary to understand evolution, while proximate causes (developmental ones) do not play an evolutionary role. Others argue instead that developmental causes are evolutionarily important and have been neglected so far.

The philosophical analysis carried out by my colleague Raphael Scholl and myself in that case splits the difference, helping to resolve that internal debate among scientists by invoking the concept of explanatory relevance (the idea that of the entire web of causality only certain portions are pertinent to explain any given phenomenon). This allows us to distinguish cases where developmental causes are required in order to understand the evolution of a structure from those cases in which developmental explanations can be backgrounded without loss of explanatory power. Some times it is the one, at other times the other.

In the case of Craver and Bechtel’s paper, instead, there really isn’t much of a discussion or disagreement among scientists who use the concepts of top-down or bottom-up causation, insofar as I’m aware. And yet, there should be! The philosophical analysis here strongly argues for the conclusion that scientists are being sloppy when they engage in either top-down or bottom-up causal talk, since all non-spooky causation can be argued to be intra-, not inter-level. Whether the scientists in question will take heed of the philosophers’ advice is, of course, an entirely different matter, which has to do much more with the sociology of interdisciplinarity than with the merits of the proposal itself.

Advertisements


Categories: Philosophy of Science

52 replies

  1. Very interesting Massimo, thanks; also sounds like an interesting workshop, and your paper too!

    Like

  2. When I try to think of these things I try to imagine simpler scenarios. For example I have a simplified computer model of the behaviour of gas particles. In the model I can observe things like compression waves, convection currents, 2nd Law of Thermodynamics, but I didn’t program any of these into it.

    All I programmed in was how a particle of such-and-such velocity changes position with time and how the momentum of particles change when they collide.

    The rest you could say was emergent behaviour or maybe constitutive effects.

    I could write an equation to describe how the compression wave moves or a differential equation to describe the rate at which one section cools. So how is there this complex behaviour in a system I wrote when I didn’t write any of that complex behaviour?

    Of course there is nothing at all mysterious about that – I can see it all happening before me and I know everything there is to know about this system.

    But there are principles which apply to collections of things that don’t necessarily apply to their components. These principles can only be understood in context of the kinds of collections to which they apply and they can be understood independently of any detailed knowledge about the components which might make up such systems.

    So the momentum of any particle in this collection is in part due to a principles which apply to the collection. For example if there are two waves approaching each other than the shape and speed of each wave will be a factor in the momentum of the particles within the wave when the waves meet.

    That seems to be a fairly reasonable account of inter level causation without any reference to anything mysterious or spooky, unless you regard the fact that there are principles that apply to collections of things.

    If I added some kind of attraction/repulsion rules to the particles then the system in general would obviously act differently. There would still be waves, convection currents, 2nd Law of Thermodynamics type behaviour, but it would be different – I would need new equations. That is bottom up causation in one sense.

    I could engineer a situation where there is another kind of bottom up causation. Suppose I set them up just so, so that the gas particles in two chambers behaved so that the hotter chamber became hotter and the colder chamber became colder. That effect would then depend sensitively on the exact momentum and position of each of the particles. Make the slightest change to just one particle and it would go back to normal 2nd law of thermodynamics type behaviour. So there is an example (contrived I grant you) of bottom up causation in another sense.

    So I apply this to the tennis player Hal. There are the physiological events described in the paper, but the precise way these events happen will depend on part on principles that only apply to collections of things and don’t apply to their components, like ambition, desire for money etc. The causal account at a physiological level depends on these. ADP is increased, but the amount by which it is increased will depend on these upper level principles. But the causal account of the that increase would be neither rigorous nor complete without an account of why it increased by that particular amount.

    That does not seem to me any more mysterious or spooky than the precise momentum and position of the particles in my computer model depending upon the shape and speed of the waves approaching each other.

    Like

  3. And let me just comment that not every person has seen every film and the enjoyment of certain films depends on the unfolding search for the explanation of some event that happens at the beginning of it. If a couple of philosophers just go and give away the ending as part of an example of causality then it will cause a number of people’s enjoyment of said film to be spoiled.

    Not considered good form for just that reason.

    Like

  4. Incidentally I am not suggesting that all higher level principles that apply to reality will happen in a computational model of the system.

    In fact I think that there are some higher level principles that happen in real systems that can never even in principle occur in a computational model of that system.

    Maybe that is what people call strong emergence.

    Like

  5. Robin,

    ‘So the momentum of any particle in this collection is in part due to a principles which apply to the collection. For example if there are two waves approaching each other than the shape and speed of each wave will be a factor in the momentum of the particles within the wave when the waves meet.”

    ” the precise way these events happen will depend on part on principles that only apply to collections of things and don’t apply to their components, like ambition, desire for money etc.”

    Yet at some level isn’t the complexity more in the modifiers than the process? As in wouldn’t ambition and desire for money be motivating the momentum of the particle otherwise known as Hal?

    Or the wave that is Hal meeting the wave that is his tennis opponent? Both focal points of ambition and desire.

    Like

  6. Of course, complexity has an energy cost as well;

    Like

  7. Robin,

    I must be particularly dense on this issue, but I still don’t get it. You keep equating rigorous with comprehensive, they are just not the same thing. The authors never talk about a comprehensive account of anything, they do talk about a rigorous concept of causation, one that excludes “spooky” stuff. Rigor does not require comprehensiveness.

    Liked by 1 person

  8. Top-down and bottom-up causality is like compiler and decompiler in computer science.

    https://en.wikipedia.org/wiki/Decompiler
    “A decompiler is a computer program that takes an executable file as input, and attempts to create a high-level source file which can be recompiled successfully.”

    Like

  9. In levels of mechanisms (see Craver 2001; forthcoming), an item X is at a
    lower level than an item S if and only if X is a component in the mechanism for
    some activity phi of S. X is a component in a mechanism if and only if it is one of
    the entities or activities organized such that S phi’s.

    Take a system S composed of a stick of dynamite and a mechanism X, a timer with an ignition. The activity of the system is “exploding”. I think it’s perfectly valid to say that X causes the explosion of S.

    Pointing out that dynamite itself is composed of certain molecules, that happen to interact in a certain way with heat is obviously correct, but on the level I’m interested in, “X causes the explosion of S” is explanatory adequate.

    And anyhow, where do you stop when you claim that one should analyse the explosion in terms of molecules and heat? A molecule also is a system, composed of a nucleus and electrons and so on. Before you know you arrive at the Standar Model. (That’s basically what I meant when I wrote that the authors seem to suggest that levels don’t exist in the context of causality).

    Like

  10. Something I forgot to mention … I get the impression that behind the discussion about levels and bottom-up causation etc. another and perhaps more complex question is hiding, namely: “what counts as explanatory adequate?”

    Liked by 3 people

  11. Some thoughts:

    “The general had a heart attack [1a] and then he died [2a]”

    “When the heart stops beating [1b], it stops transporting oxygen and nutrients to the other tissues of the body, and they cease to function [2b]”

    It’s tempting to say 2a was caused by 2b.

    But 1b caused 2b, and 2b is more of another way of saying 2 than a cause of 2.

    And I have the impression they’re saying that within levels, say biology, we can run into the same kinds of problems for example when explaining multiple ongoing processes.

    Like

  12. And I have the impression they’re saying that within levels, say biology, (…)

    Another thing I forgot to mention. Perhaps the relevance of the paper should be judged in the proper context (this context being certain discussions in biology).
    I don’t know anything about biology, so I’m not judging.

    Like

  13. Hi Massimo

    You keep equating rigorous with comprehensive, they are just not the same thing.

    Well I have tried explaining three times why I am not equating rigorous with comprehensive.

    I must be expressing myself very badly as I have not even been able to make myself understood that I am disputing that I am equating rigorous with comprehensive, since your last reply implies that I think that they are the same thing.

    Maybe we can take this one step at a time.

    You are saying, if I have understood you correctly, that a causal account that could not account for the difference in the types and nature of physiological events that occur in a high stakes tennis match and those that occur in a low pressure charity tennis match, might still be rigorous, but not complete.

    Do I understand you correctly?

    Like

  14. Corrections,

    “The general had a heart attack [1a] and then he died [2a]”

    “When the heart stops beating [1b], it stops transporting oxygen and nutrients to the other tissues of the body, and they cease to function [2b]”

    It’s tempting to say 2a was caused by 2b.

    But 1a caused 2a, and 2b is more of another way of saying 2a than a cause of 2a.

    I think the problems of top down/bottom up causation really start to show up when more complex examples are considered.

    And I think Craver and Bechtel are also saying that within levels, say biology, we can run into the same kinds of problems for example when explaining multiple ongoing processes.

    Like

  15. Hi couvent2104

    Take a system S composed of a stick of dynamite and a mechanism X, a timer with an ignition. The activity of the system is “exploding”. I think it’s perfectly valid to say that X causes the explosion of S.

    Pointing out that dynamite itself is composed of certain molecules, that happen to interact in a certain way with heat is obviously correct, but on the level I’m interested in, “X causes the explosion of S” is explanatory adequate.

    I expect that certain people might not welcome the idea that there might be a useful distinction to be made between material cause and efficient cause after all.

    Like

  16. Incidentally, I should point out that the authors of the paper do appear to be saying that they think the intentional can be explained by a sufficiently complete and complex detailing of the underlying neuro-physiological mechanism, unless I have misread them on that part.

    Like

  17. Hi Philip

    I am not sure that compilers/decompilers would be relevant as they don’t really skip levels. They just translate one symbolic scheme to another.

    Maybe it would be more relevant to consider the idea of a behavioural decompiler. This wouldn’t look at the executable at all but rebuild the source code from the behaviour of the system in operation.

    The only example of behavioural decompilers are people. We can look at a system, observe how it works and build a source code that would compile into an identically behaving system. I don’t think there are any non-human behavioural decompilers yet.

    Like

  18. Robin,

    “You are saying, if I have understood you correctly, that a causal account that could not account for the difference in the types and nature of physiological events that occur in a high stakes tennis match and those that occur in a low pressure charity tennis match, might still be rigorous, but not complete.”

    I’m saying that any causal account is never going to be complete, and yet can be rigorous because it relies on a precise account of what does and does not count as a cause (as distinct from, for instance, an internal mechanism, to use the authors’ terminology). So your question doesn’t really get to the issue at all.

    As I said, causal completeness is a chimaera, because there is a potentially near infinite number of causes for any event, but only a very small subset is going to be explanatorily relevant in any particular case.

    Liked by 1 person

  19. Robin,

    I expect that certain people might not welcome the idea that there might be a useful distinction to be made between material cause and efficient cause after all.

    I guess so … But I just wanted to give an example of bottom-up causation that – for me at least – is uncontroversial.
    One can claim it’s not really bottom-up, that the causation is intra-level (between all those molecules of the system etc.). But again: if one takes this position, where does it end? At the Standard Model? At a certain moment one has to decide what is the level of detail one is interested in (even in physics this is common). I think that on the level of my dynamite + timer, it’s uncontroversial to say that a lower level component of a system causes the activity of the entire system. And that looks like bottom-up, non-intra-level causation to me.

    Like

%d bloggers like this: