Philosophy and Cosmology: Day Three

Back for the third and final day of the Philosophy and Cosmology conference in honor of George Ellis’s birthday. I’ll have great memories of my time in Oxford, almost all of which was spent inside this lecture hall. See previous reports of Day One, Day Two.

It’s become clear along the way that I am not as accurate when I’m trying to represent philosophers as opposed to physicists; the vocabularies and concerns are just slightly different and less familiar to me. So take things with an appropriate grain of salt.

Tuesday morning: The Case for Multiverses

9:00: Bernard Carr, one of the original champions of the anthropic principle, has been instructed to talk on “How we know multiverses exist.” Not necessarily the title he would have chosen. Of course we don’t observe a multiverse directly; but we might observe it indirectly, or infer it theoretically. We should be careful to define “multiverse,” not to mention “exist.”

There certainly has been a change, even just since 2001, in the attitude of the community toward the multiverse. Quotes Frank Wilczek, who tells a parable about how multiverse advocates have gone from voices in the wilderness to prophets. That doesn’t mean the idea is right, of course.

Carr is less interested in insisting that the multiverse does exist, and more interested in defending the proposition that it might exist, and that taking it seriously is perfectly respectable science. Remember history: August Comte in 1859 scoffed at the idea we would ever know what stars were made of. Observational breakthroughs can be hard to predict. Rutherford: “Don’t let me hear anyone use the word `Universe’ in my department!” Cosmology wasn’t respectable. For what it’s worth, the idea that what we currently see is the whole universe has repeatedly been wrong.

So how do we know a multiverse exists? Maybe we could hop in a wormhole or something, but let’s not be so optimistic. There are reasons to think that multiverses exist: for example, if we find ourselves near some anthropic cutoff for certain parameters. More interesting, there could be semi-direct observational evidence — bubble collisions, or perhaps giant voids. Discovering extra dimensions would be good evidence for the theories on which the multiverse is often based.

The only direct observations that currently exists that might bear directly on multiverses is the prediction of giant voids and dark flows by Laura Mersini-Houghton and collaborators.

Carr believes that the indirect evidence from finely-tuned coupling constants is actually stronger. Existence of planets requires a very specific relationship between strength of gravity and electromagnetism, which happens to exist in the real world. There is a similar gravity/weak tuning needed to make supernovae and heavy elements. Admittedly, many physicists dislike the multiverse and find it just as unpalatable as God. But ultimately, multiverse ideas will become normal science by linking up with observations; we just don’t know how long it will take.

9:45: George Ellis follows Carr’s talk with what we’ve been waiting for a while — a strong skeptical take on the multiverse idea.

There are lots of types of multiverses: many-worlds, separated by space or time, or completely disjoint. Anthropic arguments are what make the idea go. The project is to make the apparently improbable become probable.

The very nature of the scientific enterprise is at stake: multiverse proponents are proposing that we weaken the idea of scientific proof. Science is about two things: testability and explanatory power. Is it worth giving up the former to achieve the latter?

The abstract notion of a multiverse doesn’t get you anything; you need a specific model, with a distribution of probabilities. (Does Harry Potter exist somewhere in your multiverse?) But if there is some process that generates universes, how do you test that process? Domains beyond our particle horizon are unobservable. How far should we expect to be able to extrapolate? Into a region which, in principle, we will never be able to observe.

In the good old days we accepted the Cosmological Principle, and assumed things continued uniformly forever beyond our observable horizon. Completely untestable, of course. If all the steps in the extrapolation are perfectly tenable, extrapolations are fine — but that’s not the case here. In particular, the physics of eternal inflation (gravity plus quantum field theory, Coleman-de Luccia tunneling) has never been tested. It’s unknown physics used to infer an unobservable realm. Inflation itself is not yet a well-defined theory, and not all versions of inflation are eternal. We haven’t even found a scalar field!

There is a claim that a multiverse is implied by the fine-tuning of the universe to allow life. At best a weak consistency test. Can never actually do statistical tests on the purported ensemble. Another claim is that the local universe, if it’s inside a bubble, should have a slight negative curvature — but that’s easily avoided by super-Hubble perturbations, so it’s not a strong prediction. We could, however, falsify eternal inflation by observing that we live in a “small” (topologically compact) universe. But if we don’t, it certainly doesn’t prove that eternal inflation is right. Finally, it’s true that we might someday see signatures of bubble collisions in the microwave background. But if we don’t, then what? Again, not a firm prediction.

Ultimately: explanation and testability are both important, but one shouldn’t overwhelm the other. “The multiverse theory can’t make any prediction because it can explain anything at all.” Beware! If we redefine science to accommodate the multiverse, all sorts of pseudo-science might sneak inside the tent.

There are also political/sociological issues. Orthodoxy is based on the beliefs held by elites. Consider the story of Peter Coles, who tried to claim back in the 1990’s that the matter density was only 30% of the critical density. He was threatened by a cosmological bigwig, who told him he’d be regarded as a crank if he kept it up. On a related note, we have to admit that even scientists base beliefs on philosophical agendas and rationalize after the fact. That’s often what’s going on when scientists invoke “beauty” as a criterion.

Multiverse theories invoke “a profligate excess of existential multiplicity” in order to explain a small number of features of the universe we actually see. It’s a possible explanation of fine tuning, but is not uniquely defined, is not scientifically testable, and in the end “simply postpones the ultimate metaphysical question.” Nevertheless — if we accumulated enough consistency tests, he’d be happy to eventually become convinced.

11:00: After coffee, we reconvene for a panel discussion on “the views of physicists,” featuring Pedro Ferreira, George Efstathiou, and Sean Carroll.

Ferreira talks about the paradigm shift we’ve seen over the last decade. Driven by three factors: inflation, the accelerating universe, and the string landscape. What are the alternatives? Inflation has almost no robust alternatives, unlike twenty years ago. Still not very precisely predictive, too many models. The cosmological constant is a good explanation of the accelerating universe. There are viable alternatives, but they all require severe fine tuning; the story isn’t yet finished. The landscape has been around since the 80’s, but didn’t take off until the vacuum energy became a pressing issue. It might be good if more people were digging into properties of specific compactification schemes. Is it premature to worry about the multiverse when we understand so few compactifications?

Efstathiou pulls out a white board and starts writing equations! Which I will not reproduce here (it’s basic Bayesian probability). The point is that we can only test models against other models, not in some abstract theoretical vacuum. When you don’t have well-defined alternatives, you must take what you can get. He then tells a parable about the promiscuous astronomer, which I also won’t reproduce here. (A close relative of the Sleeping Beauty puzzle well-known to philosophers.)

Carroll explains that it’s difficult to explain the low entropy of our early universe if we stick to unitary autonomous evolution of a comoving patch of space; if we don’t want to give up on conventional quantum mechanics and we don’t want to invoke ad hoc boundary conditions, we’re led directly to a multiverse. He is interrupted frequently by spontaneous applause and finishes to a standing ovation, several audience members smacking their palms to their foreheads and crying “Brilliant! Why didn’t I think of that?”

During the question period, Max Tegmark puts even money on the proposition that the Planck satellite will find evidence for an inflationary gravitational-wave background from B-mode polarization of the cosmic microwave background. Carroll quickly turns it into a formal bet, and the stakes are set at $100. Max will be paying up three or four years from now.

Tuesday afternoon: Philosophical Assessment of the Scientific Case

1:45: The afternoon session is given over to the philosophers. We begin with a panel discussion featuring John Norton, David Wallace, Wayne Myrvold, and Henrik Zinkernagel.

Norton points out that we’re all looking for ultimate explanations of everything — looking down on the idea of brute inexplicable facts. If you go down that road, you run the danger of accepting more metaphysics that you were originally planning to. Don’t get too wedded to particular criteria for distinguishing science from non-science; it leads to later regrets when you want to do something perfectly scientific that doesn’t quite fit your criterion. The problem with pseudosciences is not that they don’t make predictions; it’s that they always change to accommodate new information. “Falsifiability” is a great slogan and an irresistible sound bite, but not a reliable rule to separate science from non-science.

Wallace talks about observable vs. unobservable things. We can’t observe electrons, dinosaurs, or the interior of the Sun. We nevertheless accept them because they are non-optional parts of theories that are very tightly coupled to data. You can’t have a theory that explains dinosaur fossils without believing in dinosaurs. Multiverses come in a variety of forms — from practically useless to quite specifically useful. Not all created equal, and we need to be a bit careful.

Myrvold dislikes the very idea of a “philosophical assessment” of scientific cases; that’s a job for the people who are deeply involved with the theories and the data. Nevertheless (unsurprisingly) there is something to say. Trying to explain things at the edge of knowledge is part of the practice of science, but there is certainly some inductive risk. Hypothetico-deductive method: deduce consequences of an hypothesis and some auxiliary assumptions, go out and test these consequences, reject hypothesis if consequence is falsified. That can’t possibly be the whole story. Do we reject hypothesis or auxiliary assumptions? Quine: maybe we were just hallucinating. Process is often underdetermined.

Zinkernagel talks about the role of time in the multiverse. Could our universe be just a patch in a much older and larger structure? Depends on what you mean by “time.” We could define “supercosmic time” that orders different patches, or we could simply extrapolate from within our observable patch. General relativity allows for all sorts of time coordinates, but under special circumstances we can define a unique notion of time. To do this in the real world, you need a physical clock, and something to set the scale.

The discussion session gets into the question of policing the boundaries of respectable science, and whether the truth will eventually win out regardless of how well we do the policing. Brian Greene takes a poll: do people think that 200 years from now the truth or falsity of the multiverse (or whatever) will basically shake itself out regardless of our current investigation of what does or does not count as science? Almost everyone agrees that it will except for Julian Barbour, who doesn’t believe in time anyway (Brian’s joke).

4:00: We go right into another panel discussion, this time about “Cosmology as a Science,” featuring William Stoeger, Jean-Philippe Uzan, Jeremy Butterfield, and Huw Price.

Stoeger dives into the thicket of the demarcation problem: what is science, and what is not? It wasn’t all that long ago that cosmology itself had a doubtful status as science. There’s only one observable universe; can’t do repeatable experiments. But cosmology is not only a science, but has extended beyond the observable part of the universe. Ernan McMullin: induction moves backwards from observed effects to inferred causes. Continued success of an hypothesis leads us to conclude that “something very much like the content of the hypothesis exists in reality.” Always trying to fit our hypotheses into larger explanatory systems. A bit more optimistic than Ellis about bringing multiverses into cosmology as long as they continue to provide fruitful hypotheses.

Price has three comments for us. First, why are we asking “Is the multiverse science?” rather than simply “Is the multiverse true?”, or at least “Is it a reasonable hypothesis?” Second, there is a close analogy between the current multiverse ideas and the Steady State cosmology (details for the reader to work out). Third, think about the notion of “explanatory relief.” The inflationary multiverse provides relief for the puzzle of why the universe allows life; but we don’t think that the Everett multiverse provides relief for the puzzle of why some particular person exists. The Lewisean multiverse (all logically possible worlds exist) provides relief for all sorts of questions — but what good is that? There’s a slippery slope here, trading off explanatory power for triviality. TIme for a fourth comment: In everyday life, we explain the future in terms of the past (asymmetric in time). But what is the justification for that in the context of cosmology?

Uzan wants to talk about cosmology with a lower-case “c,” what we call “physical cosmology.” Standard cosmological model relies on assumptions about gravity, matter, symmetry, and global structure of the universe. All should be subjected to empirical tests. For general relativity, it would be nice to have more tests on cosmological scales (low acceleration, low curvature). For example, using various forms of large-scale structure data — weak lensing, galaxy maps, velocities, integrated Sachs-Wolfe effect. We can also imagine testing the assumed symmetries of space, as embodied in the Copernican principle. With precise spectra, we could measure the change in the redshift of an object over time. And of course, we can look for a compact topology to the universe.

Butterfield wants to keep his remarks short so that he doesn’t say anything incriminating that ends up on this blog. [Ed. note: everyone knows about the live-blogging and there is a slight undercurrent of distrust, although Jeremy was just joking. I think.] He congratulates the cosmology community in general, and George Ellis in particular, on their endogenous amphetamines. Wants to encourage plucky theories being studied by small groups of researchers (“weeds in the garden,” in a nomenclature from a previous session). How much do/should controversies in the foundations of statistics interact with debates in fundamental cosmology? In particular, how do they impact the measure problem?

In discussion, George Ellis quotes G.K. Chesterton: “the only people who have an explanation for everything are madmen.”

Tuesday evening: Why is there Something Rather than Nothing?

9:30: The conference closes with a posh dinner at Balliol College, featuring a talk by John Hawthorne on the primordial existential question. Then we have a response by Max Tegmark. Apparently my principled refusal to go to talks after 9:00 p.m. is no more reliable than my refusal to attend meetings that receive funding from the Templeton Foundation.

But I didn’t actually bring my laptop to the talks, so the summaries will be short and sweet. Hawthorne didn’t offer an answer to the question, but tried to provide a roadmap to possible answers. He distinguished between different notions of “nothing,” from true non-existence to an empty and involving universe. Then he distinguished between explanations from law — something exists because there is a deep principle of nature which, perhaps among other things, requires something to exist — and explanations from probability — there is a measure on the space of things that might possibly exist, and the measure assigned to the empty set is zero or at any rate small. Long story short, metaphysicians have not reached a consensus on this question.

Tegmark decided to tackle the question of what kinds of things there could be, rather than why at least one of them was real. He has a classification of multiverses: Type I is just a really big space that is pretty similar throughout, Type II is an inflationary kind of multiverse with distinct (although perhaps connected) bubbles, Type III is different branches of the wave function. His favorite option is the Type IV, in which all possible mathematical structures exist; that’s closely related to David Lewis’s ideas about modal realism.

I had a question I wanted to ask, as follows. Imagine that we agreed that any universe could be described by some sort of mathematical structure; perhaps a wave function evolving in Hilbert space, or a discrete cellular automaton, or a single point, or even the empty set (“nothing”). The question is, what kind of possible answer to the question “Why is the actual universe this structure rather than that structure?” could one imagine providing that would be more fruitful, give us more information, than a simple statement of the brute fact of which structure it was? Isn’t it necessarily like trying to discern a pattern from a single event?

But I was sensible enough to recognize that going to sleep was a higher priority than figuring out the meaning of existence. Maybe some other time.