The cosmological “multiverse” model talks about regions far beyond the observable portion of our universe (set by the finite light-travel distance given the finite time since the Big Bang). Critics thus complain that it is “unfalsifiable”, and so not science. Indeed, philosopher Massimo Pigliucci states that instead: “… the notion of a multiverse should be classed as scientifically-informed metaphysics”.
Sean Carroll has recently posted an article defending the multiverse as scientific (arXiv paper; blog post). We’re discussing here the cosmological multiverse — the term “multiverse” is also used for concepts arising from string theory and from the many-worlds interpretation of quantum mechanics, but the arguments for and against those are rather different.
Any theory of cosmology supposes that the universe continues beyond our observable horizon (why wouldn’t it?, since that horizon is an accident of our current location in time and space, and an “edge” would be far harder to explain). From there, one can adopt a “more of the same” model beyond the horizon, or one can accept arguments that physical conditions might be radically different in different, very-remote regions, thus producing a “multiverse”.
The job of science is to model the world around us, a reality that we learn about by empirical means. The “scientific method” consists of constructing models of reality, comparing them to empirical reality, and then updating them to work better, in an endless iteration.
One oft-made claim that is not true is that science is limited by necessary metaphysical assumptions (one posited example being philosophical naturalism, an a priori rejection of the possibility of gods). Thus one should be wary of philosophers declaring that some investigation violates a necessary condition of science, and so is “not science”.
It may well be that there actually is a multiverse (distant regions of space having very different physical conditions) and so science, which is our best attempt to describe the universe, should be able to suggest that possibility. It should not be declared unscientific by philosophical fiat. Our task is to adjust science to match nature, not put artificial limits on it.
And so to “falsifiability”, sometimes claimed by the “Popperazzi” as a necessary criterion for being scientific, after Karl Popper proposed it as a way of distinguishing science from what he saw as pseudo-sciences such as Freudian psychoanalysis and Marxist social criticism.
Falsification does not work if over-interpreted as a simplistic criterion. Cosmologist Peter Coles remarks: “I’ve never taken seriously any of the numerous critiques of the multiverse idea based on the Popperian criterion of falsifiability because … that falsifiability has very little to do with the way science operates”.
But interpreted properly there is a large measure of truth to it. If some claim has no relevance to anything observable then it is not “about” the empirical world, and so is not part of what science is trying to achieve. Where ideas are carefully designed to avoid any possibility of being refuted the practitioners are not pursuing science, in that they are not attempting to match reality, and instead they are attempting to bolster an ideology. A classic example is the idea that “God always answers prayers, but sometimes he says no”.
Thus theologians (to give one example of a pseudo-science) will continually adjust their theology to avoid saying anything definite and testable. If preservation of a prior commitment is given a higher priority than matching reality, then one is doing pseudoscience. In contrast, a scientist will (or at least should!) try hard to make falsifiable predictions, looking for ways to test and refute their models, since that iteration leads to their improvement. That difference in attitude is the demarcation that Popper was pointing out.
But that is not the same as requiring immediate potential for falsification as a criterion for doing science. Science is pragmatic, a matter of doing ones best. We will always be hitting practical limits on what we can do and science is always limited and fallible. We can never have full confidence in any statement about what is beyond the observable horizon, but then we can never have full confidence in any part of science. The point is that, so long as we’re doing our best, we’re not being unscientific. That holds even where “our best” is pretty limited.
The important point is that it is not necessary to be able to test all aspects of a model, one merely need test the model as best one can. For example, one can use a Solar System model to predict the occurrence of an eclipse next year, and then test it. But the same model could predict eclipses millions of years ago, and there would be no conceivable way of directly testing those predictions. Indeed all scientific models are like this, where we can, in practice, test some but not all predictions of a model.
I doubt if anyone using a Solar System model to generate predictions of eclipses thirty millennia ago would be accused of being “unscientific”. The eclipse-prediction model would be testable only within a narrow time-span, but having tested it within that time span we could legitimately and scientifically have confidence in its predictions over a much wider time span.
A cosmological multiverse model is similar. It attempts to model all space, both within our observable universe and outside it, and we can test the predictions it makes about space within our observable horizon. If it does well there, we can then legitimately have some degree of confidence about its predictions for outside our observable horizon. Nothing about that is unscientific.
If the multiverse model had been proposed to apply only to regions beyond the observable horizon, with no relevance or implications for anything within it, then the accusation that it is unscientific would be fair, but that is not the case.
The concept of a multiverse is a prediction of our current best attempts to explain cosmology within our observable horizon. In order to make the Big Bang model work, and match empirical data about our early universe, one needs to invoke “inflation”, a rapid exponential expansion of the universe during its first tiny fraction of a second. The usual way this is modelled is by an “inflaton field” energy density which would, given the known physics of General Relativity, drive an exponential expansion.
The physics of the inflaton field are not well understood, and the whole inflationary model is unproven, but the point relevant to this post is that the scenario is constructed in order to model empirical data (the smoothness and uniformity of the microwave background; the fact that the universe’s density is so close to the critical density; the absence of particles such as magnetic monopoles; et cetera) and thus is eminently a scientific model. It makes predictions that are testable, such as a signature of gravitational waves in the microwave background.
If the inflationary scenario is correct, then our universe must have dropped out of the inflationary state by a quantum fluctuation. But, quantum fluctuations are localised in space. Therefore, only a limited spatial region would drop out of the inflationary state owing to that fluctuation. The inflaton field would continue driving exponential expansion elsewhere. And, if the quantum fluctuation can happen here, then it can happen elsewhere; and thus almost inevitably one would have a large number of pocket “normal state” universes (one of which would contain us) strewn within a vastly vaster inflationary-state space (in much the same way that holes occur within an Emmental cheese).
The physical conditions in the inflationary-state regions would be very different from those in the numerous “normal state” pockets, and hence there would be a multiverse. The different pocket universes are then simply separated by vast distances.
It is very difficult to have an inflationary model of Big Bang cosmology without having a multiverse, and it is currently hard to explain the observed characteristics of our observable universe without invoking an inflationary cosmology. Thus, our current best scientific models tell us that we live in a multiverse. That is a scientific conclusion and we should place some credence in it, though only some credence since inflationary physics is still poorly understood.
That is no different, in principle, from having confidence in the prediction of eclipses in the distant past, even though we can’t directly observe them, where that confidence arises because they are predicted by a model that has been tested and validated when applied to eclipses that we can observe. Regardless of whether the inflationary/multiverse variant of Big Bang cosmology turns out to be true, it is certainly a model within the scope of science that has been adopted and developed for scientific reasons.
Thus I think that Peter Woit is missing the point when he says: “the problem with the multiverse is that it’s an empty idea, predicting nothing”. The cosmological multiverse is not invoked to try to generate novel predictions, it’s invoked as a necessary implication and consequence of models developed and tested to explain the parts of the universe that we can indeed see.
We could falsify the predictions of ancient eclipses by showing that the model that predicted them does not work when applied to eclipses that we can observe, and we can, in principle, falsify inflationary/multiverse predictions by showing that inflationary/multiverse models do a poor job when compared to data from within our observable horizon (a lack of gravitational-wave signatures in the microwave background might indeed do that).
A naive Popperazzi (going beyond anything Popper himself said) might insist that science must be strictly limited to direct observables and immediately falsifiable claims. They might then insist that only claims relating to space within our observable horizon are “scientific” and thus, if we extrapolate the model beyond that, then we are doing “metaphysics” or something.
That would imply that eclipse predictions relating to the time-frame of recorded history are “scientific”, but that predictions relating to a time before that would be metaphysical (though the boundary would in any case be fuzzy, since who knows what cave paintings relating to an ancient eclipse might be found?). That would strike me as a fairly pointless and merely semantical distinction, based on an overly narrow conception of science. You’d be forced to conclude that an eclipse occurring a decade ago in Antarctica and viewed by no-one was also “metaphysical”. And likewise for an eclipse, visible from New York, predicted for 2099 (or pick whatever time into the future makes it insufficiently falsifiable in your eyes). Some philosophers want to limit science narrowly to empirical data, and deny it the wider conceptualising about the data and what they imply, but science has always been just as much about concepts as about data; without concepts, accumulating data would be mere “stamp collecting”.
Overall, I can only agree with Sean Carroll:
I argue that the way we evaluate multiverse models is precisely the same as the way we evaluate any other models, on the basis of abduction, Bayesian inference, and empirical success. There is no scientifically respectable way to do cosmology without taking into account different possibilities for what the universe might be like outside our horizon. Multiverse theories are utterly conventionally scientific, even if evaluating them can be difficult in practice.
Agreed. A theory that remains consistent with observation and/or supplies a tentative basis for explaining puzzling aspects of other theoretical abstractions or models that ARE confirmed by observation or experiment is a legitimate example of the scientific process.
You’re mostly doing the same thing Carroll is, invoking straw man naive arguments just in order to knock them down. When you quote me and say I am “missing the point”:
“Thus I think that Peter Woit is missing the point when he says: “the problem with the multiverse is that it’s an empty idea, predicting nothing”. The cosmological multiverse is not invoked to try to generate novel predictions, it’s invoked as a necessary implication and consequence of models developed and tested to explain the parts of the universe that we can indeed see. ”
you missed the next sentence, which is about exactly what you say I am missing. That next sentence, explaining the one you do quote, is:
“It is functioning not as what we would like from science, a testable explanation, but as an untestable excuse for not being able to predict anything.”
This is the whole reason I got involved in criticizing the multiverse business, precisely because it is claimed to be an implication of string theory and being used to justify its failure (string theory implies string landscape implies multiverse of different string vacua with different physics, so no way string theory can predict anything testable). See for instance
One of the main points I was making in the blog posting and Inference piece is that you aren’t saying anything scientific about a multiverse unless you specify what theory/model you are talking about (which Carroll doesn’t do). “String theory” is not a well-defined enough theory to have any non-empty implications about any multiverse it supposedly implies. Models of a supposed inflaton field are a different and much more complicated story. Of course I’m aware of the argument “a class of inflaton field models for which we think we have some evidence also exhibits eternal inflation and thus will produce a multiverse”. After a long argument about the strength of evidence for various features of specific inflaton field models perhaps you will get some evidence implying an eternal inflation type multiverse. But this kind of multiverse will have pretty much the same physics in every universe, and is not the kind of multiverse being heavily promoted by Carroll and others, which is supposed to have different physics in each universe and explain why string theory predicts nothing.
Unfortunately there is no standard terminology to distinguish different multiverse models, so you need to determine from context what someone is talking about. The context for what I wrote was that it was discussing Ellis’s quote about the “anything that can happen does happen in most multiverses” problem, so it should have been clear that I was talking about multiverse models with that feature.
Hi Peter, I’m staying out of the whole string theory side of things, so was only really defending the most “vanilla” multiverse — eternal inflation in which all the bubble universes might have the same physics. If you have inflation, then it seems very hard not to have eternal inflation along with pocket universes, so that flavour of multiverse doesn’t seem at all outlandish to me. Indeed, it doesn’t seem any worse in philosophical/falsifiability terms than simply extending our observable universe to infinity, as standard cosmological models have usually done.
Whether the different bubbles have different physical constants and different physics seems to me much more speculative again than eternal inflation.
By the way, on the Ellis quote that “anything that can happen does happen in most multiverses”, surely that would also be the case in a vanilla same-physics multiverse, and indeed would also be the case in any model where space simply extends to infinity? There would simply be a wider range of what “can happen” in the more exotic types of multiverse with different physics.
Pingback: Beyond Falsifiability | Not Even Wrong
Professor Carroll (and other physicists who favour the inflationary multiverse model), are simply not defending or discussing the kind of multiverse which arises purely from a single inflation field and leads to the exact same physical constants in each of the causally disconnected “patches”.
I don’t know of anyone who finds this notion in the least bit controversial – the consideration that there are likely regions of the universe beyond the particle horizon which we can’t observe but where the same underlying physics applies. This type of multiverse is not only uncontroversial – it doesn’t aid any arguments, whereas the string landscape multiverse (with the different vacua and constants) is the one that can be utilized as a convenient excuse for our inability to calculate the observed values of the cosmological constant and the higgs mass – the so-called “fine-tuning” issues.
This multiverse model relies on string theory and eternal inflation.
I was indeed trying to defend the easiest case to defend, a cosmological multiverse without asserting that the different bubble universes have different physics or different physical constants. That might be fairly mundane and accepted as “scientific” to cosmologists, but I’m not so sure that philosophers would accept it with equanimity.
I can’t be completely sure what Ellis had in mind, but his reference to multiverse proposals with “no unique observable output” seemed to me to refer to multiverse models that produce arbitrary physical laws, not just arbitrary states compatible with a specified set of physical laws (by “physical law” I mean e.g. the Standard Model with its choice of fundamental parameters). I may be wrong though.
As for the idea that whatever physics is behind the big bang, it could also have caused lots of other big bangs producing other copies of our universe with the same physics, sure that’s possible, and as you note, not that different than assuming lots more universe of the same kind beyond what we can observe. Carroll though is specifically writing about multiverse proposals that “[affect] how we understand what we observe.” Besides the string theory landscape, which is behind much of the interest in multiverse models, note that Carroll’s first book was about his proposal that the multiverse explains the arrow of time problem.
Your interpretation of Ellis’ stance is spot on. Consider this statement of his from the following 2014 article in Scientific American:
Why the Multiverse May Be the Most Dangerous Idea in Physics
Proof of parallel universes radically different from our own may still lie beyond the domain of science
“Astronomers are able to see out to a distance of about 42 billion light-years, our cosmic visual horizon. We have no reason to suspect the universe stops there. Beyond it could be many—even infinitely many—domains much like the one we see. Each has a different initial distribution of matter, but the same laws of physics operate in all. Nearly all cosmologists today (including me) accept this type of multiverse, which Max Tegmark calls “level 1.”
Yet some go further. They suggest completely different kinds of universes, with different physics, different histories, maybe different numbers of spatial dimensions.”…”
And Professor Ellis is, of course, perfectly correct in his apprehensions.
Have Sean stand at the edge of our universe and shine a flashlight beyond it. If it emits light then we have a multiverse! Until then it’s just a theory!
This is not my area, but what you wrote here makes a lot of sense to me. Like others here, however, I had read previous writings by Carroll as arguing for many-worlds, a very different proposition. I distinctly remember a blog post by him a few years back whose argument appeared to reduce to “it must be true because the math checks out”, to which I would reply that it is also the practice of science to carefully go back over one’s assumptions and models if they have lead to a patently absurd conclusion. I mean, if a fellow biologist’s analysis ‘shows’ that a genus of flowering plants originated in the Precambrian before the first fossils of multi-cellular life then I’d also expect them to figure out what they did wrong instead of rubbing their hands in expectation of a Nature paper.
(While on this topic, in the case of many-worlds it would be great if somebody in the know could clarify what its proponents actually claim. It sometimes seems like a bait and switch where the claim is either a sensationalist and rather un-thermodynamic “every event duplicates the entire universe down to the last particle in a trousers-of-time fashion” or “that is not what I meant, the universe doesn’t get duplicated, stuff only happens in Hilbert space; you ask what that means?, sorry, gotta run”, depending on whether the proponent has just been challenged or not.)
Yes, that’s true, though the QM “many worlds” multiverse is a pretty different issue from the cosmological multiverse.
In brief, a cosmological multiverse has different universes simply separated in space; a string-theory multiverse has different “universes” residing in different dimensions of a multi-dimensional super “space”; and the QM many-worlds multiverse has the different universes being different “regions” of the QM wavefunction.
I don’t claim to understand it, but as far as I can make out, QM many-worlds advocates argue that:
The wavefunction itself is ontological, so what actually “exists” is not space or particles but the mathematical structure of the wavefunction. There is one wavefunction for the entire universe/multiverse. That wavefunction then decoheres into different “terms” of the wavefunction, and what gets “duplicated” is different terms of the wavefunction. The different “worlds” are then not separated in space, and nor is *space* duplicated, but are simply different “regions” of the mathematical wavefunction.
This whole conception seems distinctly weird to me, but then maybe I’m misunderstanding it.
This whole conception seems distinctly weird to me, but then maybe I’m misunderstanding it.
Glad that I’m not the only one!
Thanks for taking time for that stab at an explanation. It is the clearest I have seen so far, and I note that it was not formulated by a proponent of the idea…
It seems like people are still conflating a few different issues in this discussion. Sticking with the cosmological multiverse for now, in *any* model being discussed the same underlying physics holds throughout the universe. The issue is that the same underlying physics can lead to locally different outcomes. The classic example is a magnet, where, as it cools down, local patches take on a preferred direction as all the microscopic generators of the magnetic field near each other line up. So different regions have a differently oriented north and south poles, while the underlying theory doesn’t have any preferred direction. (For a magnet the differences ultimately arise from tiny non-uniformities in the structure, edge conditions, etc., but the point is that the magnet becomes infinitely sensitive to such tiny effects under the right circumstances, so it can’t remain in the “unbroken” state where every microscopic piece is randomly oriented and there are no macroscopic domains with a preferred direction.)
Anyhow, the analogous idea is that during inflation different regions of the universe will also go through a phase transition that can result in different “macroscopic” conditions in different patches. The simplest case would be that different regions stop inflating at differing times as Coel describes. However, we don’t know the correct underlying theory that drives inflation or other complications that may describe a more fundamental theory. The Standard Model(SM) of particle physics is already a theory of symmetry breaking like the magnet described above. But it is a fairly simple breaking. It may be, however, that the SM itself is just a low energy effective theory resulting from a larger theory that undergoes an analogous process. There are reasons to think this is true, but it certainly isn’t proven at this point. If true though, then there could be different “Standard Models” realized in different regions of space. The underlying physics would be the same everywhere, but the apparent physics in low energy science (meaning almost anything we or some other civilization could do) would be different. Speaking loosely, the bigger the underlying theory, the more different local possibilities exist.
So, the extreme version of this is string theory, where the fundamental theory seems to allow an astronomical number of different realizations that could pop up as the universe cools in different regions. String theory was not designed to do this, it just looks like a consequence. The question then is whether we have sufficient reason to believe in string theory, or some other less ambitious possibility that still gives us multiple possible local solutions. If yes, then it’s not at all unscientific to talk about whatever multiverse possibilities the theory entails. At the moment, I don’t think anyone can argue compellingly that we must believe any specific theory beyond the SM. (Although we know that something is needed to explain inflation, dark matter, the baryon asymmetry, etc.) So Carrol is correct that the idea of a multiverse is scientific and this is true generically, the basic principle is the same in a string universe as in a vanilla SM with some simple inflation model. But we can’t justify believing in the more complicated multiverse unless we can justify the models that give rise to them and I would say we are pretty far from that at the moment.
A much more fraught proposal, which I think is the root of what some people are objecting to, involves invoking the stringy multiverse for anthropic reasons. That is, some people claim that various coincidence questions, e.g., why is the cosmological constant so small, can be answered if there is a vast array of possible local universes and observers like ourselves can only arise in those which satisfy the “coincidence” being questioned. Trying to use this as evidence for a big multiverse seems extremely dubious to me. If we “knew” a big multiverse existed it would be an important point to make, but I don’t think one can reverse it to say that the multiverse is the only explanation. I don’t think it is even well-posed what we mean by an explanation in this scenario.
Whether or not you agree with him, those confused about different types of multiverse should read Max Tegmark’s Our Mathematical Universe (which I reviewed for The Observatory) or his papers on which the corresponding parts of the book are based.
Max’s nomenclature is somewhat unorthodox, but he uses it consistently and it has its own logic.
For example, his Level I multiverse is not what most people would call a multiverse at all, but rather just the universe (with Max’s universe being the observable universe, i.e. that within the particle horizon). To first order, the various multiverses are different concepts, but Max and Anthony Aguirre have a punny paper where they try to combine two of them. Many aspects of the multiverse actually appeared in Olaf Stapledon’s science fiction.
This article about the multiverse is interesting too.