Skip to content

Causation of Everything

January 28, 2014

We’ve had causation come up a few times on the blog before (particularly in Mike’s discussion of miracles). For this post, I want to raise some questions about what to say when causation get really big—when we start talking about the state of everything at one moment in time causing everything at the next—and whether such talk is sensible. Such questions are particularly relevant in the context of cosmology.

Usually when we make causal claims or explanations, we’re talking about a local causation: a particular event (or event-type, or state of affairs) that occurs in a relative small finite region of space-time, causing an event at another. We might speak of the high-pitched tone causing the glass to break, or CO2 emissions causing increased polar ice melting. While we might have some difficulty in identifying very diverse and diffuse causes and effects, we presume that the causes and effects are still local.

But what about system-wide causal claims? Assuming that a set of billiard balls, for example, constitutes an effectively closed system, could we claim that the entire configuration of billiard balls at one time causes the entire configuration at the next moment in time? Or could we claim that the state of the entire universe at one time causes the state of the universe at a later time? (Note that I’ll be leaving ‘a moment in time’ as a loose and vague notion here—whatever account of causation we use will have to be compatible with general relativity, but I won’t go into how that might be done here.)

In ‘On the Notion of Cause’ (1912−3), Russell famously argued that we should jettison the notion of causation altogether. His main concern was that, given the global laws we have in fundamental physics, nothing less than the entire state of the system at a given time would be enough (given the laws) to necessitate any event in the next. So as far as we think causation requires nomic necessitation, we would need to consider the entire state of the system as a cause. And he took this to be a reductio of the position: if causes were also to be general types of events, of the type science could investigate, there could be no cause-based science. Here’s Russell:

In order to be sure of the expected effect, we must know that there is nothing in the environment to interfere with it. But this means that the supposed cause is not, by itself, adequate to insure the effect. And as soon as we include the environment, the probability of repetition is diminished, until at last, when the whole environment is included, the probability of repetition becomes almost nil. (Russell 1912−13, pp. 7−8)

Either the causes would be so extensive and detailed as to be unique, and not the subject of scientific investigation, or they would not necessitate their effects. Science has to study repeatable events, and system-wide states would never be repeatable in the way required.

So, in Russell’s work we actually have an argument in favour of system-wide causation: it allows us to take the causal relation to be necessitation by fundamental laws. But we also have an argument against system-wide causation: it isn’t about the kind of repeatable events that science is concerned with. The argument in favour of system-wide causation seems clear enough, but what are we to make of the argument against? It seems that cosmology is precisely a field that is interested in non-repeatable events. Perhaps cosmology does not describe events at a sufficiently fine-grained level to explain how many actual events are nomically necessitated, but what about large-scale phenomena? Surely cosmology aims to account for those?

However, some later developments that followed Russell’s work didn’t advocate for system-wide causation, but dropped the requirement that causes had to nomically necessitate their effects. Instead, we were to explain what was going in causation using counterfactuals and notions of intervention. Claiming a causes b is roughly to claim that had we intervened on a in a suitably surgical of way, this would have been a way of influencing b. The interventions themselves were characterised as causal processes, as processes that that disrupt some of the causal chains already present in the system, while leaving others intact, and so allows us to test what causal chains there are. The main expositors of this approach are Woodward (2003) and Pearl (2000). These interventionist approaches don’t attempt to reduce causation to something else, but instead offer an elucidation of various causal notions in terms of other ones.

However, under such the interventionist approach, it’s hard to see how we can talk about system-wide causation. Interventions were envisaged as process that originated from outside the system we were studying. How is this approach to work when there is nothing outside the system? Here is Pearl on the issue:

…scientists rarely consider the entirety of the universe as an object of investigation. In most cases the scientist carves a piece from the universe and proclaims that piece in – namely, the focus of investigation. The rest of the universe is then considered out or background and is summarized by what we call boundary conditions. This choice of ins and outs creates asymmetry in the way we look at things, and it is this asymmetry that permits us to talk about “outside intervention” and hence about causality and cause-effect directionality. (Pearl 2000, p. 350).


‘If you wish to include the entire universe in the model, causality disappears because interventions disappear – the manipulator and the manipulated loose their distinction. (Pearl 2000, p. 349−50).

We might claim this as a feature of the interventionist approach: the approach makes it clear how the causal structure of the world is tied to a particular limited perspective we take on it as experimenter or intervener, dividing the world up, and is not something that can be understood outside of these divisions. Are we then content to cease dealing with causal notions in sciences like cosmology?

There are a few remaining options on the table that I’ll note briefly. One is to start with the interventionist framework, but then extend our causal notions so that they can be system-wide. We might, for example, attempt to reduce the interventionist counterfactuals to law-based ones that can then be applied to whole systems—I take Albert (2000) and Lower (2007) to follow this route. Or we might keep causation as a primitive relation that holds between local events, and build up system-wide causes out of those. Another quite different option is to go pluralist about the notion of causation: perhaps we were wrong to think that a single notion applied to all contexts. We should keep nomic necessitation as what counts for cosmology, and intervention as what counts for other contexts.

Whatever option we take here, the case of cosmology seems a useful testing ground for accounts of causation and their commitments regarding global causes.


Albert, David Z. 2000. Time and Chance. Cambridge, Mass.: Harvard University Press.

Loewer, Barry. 2007. Counterfactuals and the Second Law. in Causation, Physics, and the Constitution of Reality, ed. Huw Price and Richard Corry, 293−326. Oxford: Oxford University Press.

Pearl, Judea. 2000. Causality. New York: Cambridge University Press.

Russell, Bertrand. 1912−13. On the Notion of Cause. Proceedings of the Aristotelian Society, New Series. 13: 1-26.

Woodward, James. 2003. Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press.

4 Comments leave one →
  1. MIlkshake permalink
    January 28, 2014 3:01 pm

    Could there be a situation in which A causes B and B causes A?

  2. Hans permalink
    January 29, 2014 9:31 am

    You dismissed relativity near the beginning of the article as something to be dealt with later, but it seems part and parcel of the whole discussion to me. In particular, when you suggest “Or we might keep causation as a primitive relation that holds between local events, and build up system-wide causes out of those,” that idea of locality seems in keeping with the finite speed-of-light (or speed-of-information). There is actually a finite boundary on what events can influence something happening here and now, the (past) light cone, which starts to define locality for you.

    I don’t think this changes the argument about interventionist definitions of causality (since the experimenter is of necessity within the light-cone of the system being experimented on). Interventionist causality is a satisfying description of how experiments are done, but it doesn’t fail just because there is no experimenter – unless you believe that a tree falling in the forest doesn’t make any sound – or because the experimenter is part of the system – since the experimenter is always part of the system, and just has to be careful about it. Anyway, in cosmology the problem isn’t that we’re in the system, it’s that we can’t perturb the system: all we can do is observe, model, and constrain.

    I would think that ultimately an atomistic (local) definition of causality would be the most satisfying – starting very locally with “what caused this particle to change direction?” “it exchanged a photon with this other particle” and so-on backwards in time. This effectively would reduce causality to a history of a system’s past, in which the laws of physics appear as trends. Experimentalists are able to ‘replicate history’ locally (or at least, the ‘important parts’ given the resolution of the experiment) and study it with slightly different inputs, to collect information on those trends.

    Then the study of causality probably fails at the CMB (maybe a bit earlier, as there’s imprints of earlier history on the CMB radiation). It would be weird (in the sense of unsatisfying) to claim that causality itself started there, or at any point within a reasonable model that explains the CMB: if there’s a reasonable model and it depends on a ‘first cause’, then at that point only would causality fail.

  3. February 1, 2014 2:31 pm

    Hi Alison,

    Interesting stuff. I’m interested in trying to extend the interventionist account, and one result (benefit?) of that would be to allow for interventionist causation of everything. Interventionists do usually impose the constraint on causal networks that it should be physically possible to intervene on each variable, but we can consider relaxing that (at least in contexts when we’re talking about system-wide causation). The results aren’t obviously absurd – we just need to suppose that there exist the corresponding causal networks. At least some of the interventionist counterfactuals involved will end up being counter-nomic, and perhaps that will cause some complications. But most people do seem to think counter-nomic conditionals still have reasonably determinate truth-values.

    That all assumes fundamental physical laws are metaphysically contingent. If we think they’re necessary, then to get the same result we have to loosen the restrictions still further, so that it doesn’t even have to be metaphysically possible to intervene on each variable in our models. Then we get lots of weird results; all sorts of dependencies like part-whole come out as causal, and lots of the interventionist counterfactuals have metaphysically impossible antecedents. (I have a draft on my website which explores this approach in the context of grounding.)

    all best

    • Alison Fernandes permalink
      February 7, 2014 7:28 am

      Hi Al,

      Thanks for the interesting comments. I guess I wasn’t thinking that it must be physically possible to intervene on each variable under the interventionist account (or at least under some interpretation of physical possibility). In a sense, what interventions can be considered are set by the causal modelling itself, and the connection to system-wide laws is a little obscure. But I agree your approach goes further than the generic interventionist one I had in mind.

      I guess my concern would be justifying why we should think of these further networks you have in mind as causal. Even if we accept conditionals and counterfactuals associated with them, that doesn’t strike me as enough.But of course, that doesn’t get in the way of using interventionist modelling for all sorts of other purposes too- which is what I take it is a large part of what you had in mind.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: