On November 11, Sabine Hossenfelder, a theoretical physicist and well-known science communicator, posted the following tweets, and started an incredible shitstorm among both philosophers and physicists:
This is the sort of thing that scientists will happily say, and which will cause philosophers of science to fly into a murderous rage. This isn’t because the scientist’s claim is false, exactly, but because it’s almost certain to be conceptually confused sophistry unless you are very careful. And scientists are rarely sufficiently careful.
Philosophers of science will have three huge objections to Hossenfelder’s first tweet. First: this is not a scientific claim, it’s a philosophical claim. There is no experiment in the world that you can run which would establish whether or not a claim or a theory is “scientific.” Hossenfelder is assuming, and then applying, a theory of what makes a claim “scientific,” and theories of this kind are theories in the philosophy of science. Hossenfelder is an excellent physicist. The guy she’s dunking on, Philip Goff, is an excellent philosopher of science. So Hossenfelder’s not playing on her home turf here. That’s not to say that she’s wrong, and scientists may well have insights into their own practices that philosophers of science miss. But philosophers of science generally read scientists when those scientists opine about philosophy of science, and many philosophers of science began their careers as scientists. So while it’s absolutely fair to request that philosophers of science have an appropriate intellectual humility in debates with scientists, we philosophers should be entitled to similar humility and courtesy.
The second two, substantive objections, will have to do with the content of what Hossenfelder is saying. First, she’s assuming a certain account of what makes a claim “scientific.” This has been a subject of extreme controversy in the philosophy of science for about a century. Anything that you say about what makes a claim “scientific” or not will face a mess of well-known problems and objections. Hossenfelder isn’t specific about what her criterion of scientific-ness is, but it’s clear that it has something to do with observation: if something is unobservable, then claims about it aren’t scientific. And this brings us to the big problem: what does it take for something to be observable? Again, this is a huge controversy in the philosophy of science and Hossenfelder’s invocation of this concept is what kicked off our controversy.
So let’s back up a second and look at the controversy in the philosophy of science over “observables.” This arises from a larger controversy: the debate between realists and instrumentalists. Realists say that our scientific theories are in the business of correctly representing the way the world is. “Science aims at truth” is the common slogan, although it’s somewhat unclear what it might mean for something like science to “aim at” anything. Instrumentalists, on the other hand, think that science has nothing to do with truth. Scientific theories are a tool for prediction (predicting what we’ll see in various observations and experiments) and control (building technology). So long as our theories do a good job of prediction and control, they’re good theories, and it doesn’t matter whether or not they’re true. In fact, it may not even make sense to talk about a “true” scientific theory at all. Theories are good when they work, bad when they don’t, and truth doesn’t enter into the equation anywhere.
One big problem for instrumentalists is to illustrate the extent of their instrumentalism. So, for instance: are there tables and chairs? I’m sitting on a chair at a table; will instrumentalists admit that that is true? Or will they say that “I’m sitting on a chair at a table” is just the sort of claim that is useful for prediction and control? Presumably, instrumentalists will say that claims like that can be genuinely true. Instrumentalists aren’t global skeptics, they aren’t swayed by Cartesian worries that we might be brains in vats or deceived by evil demons. They just don’t think that scientific theories are in the truth business, although other claims are. That’s fine as far as it goes. But then this just raises a question: what sorts of claims should we be instrumentalists about, and what sorts of claims should we be realists about?
The usual answer to that question appeals to the notion of observation. If something is observable, then we should be realists about it. It exists, and we can refer to it and speak truly when we do so. But if something is unobservable, then we shouldn’t be realists about it. Things like electrons are too small to be observable, and thus theories about electrons aren’t in the truth business. They’re merely useful theories that work or not. And it turns out that electron theories have been very useful. So we accept those theories, but that doesn’t mean that electrons exist or that those theories are true. Electrons are the paradigm example of an unobservable entity that features in successful theories.
With all of this in mind, go back to the last sentence in Hossenfelder’s tweets: “In a nutshell, science cannot make any statements about the existence of something we can’t observe, may that be gods or other universes.” Electrons are the paradigm case within the philosophy of science of an “unobservable” entity. So… is Hossenfelder saying that electrons don’t exist? Goff put precisely that question to Hossenfelder.
And then philosophers and physicists spent a week yelling at one another about realism, instrumentalism, whether electrons exist, and who is being arrogant (hint: everyone). Hossenfelder summarized the state of play a couple days later:
So, who’s right? Fortunately, I’m here to tell you. Good news: This is a huge misunderstanding that’s pretty easy to clear up. Bad news: The whole debate, it turns out, was very stupid.
There are two issues here. One is that Hossenfelder probably misspoke somewhere. She said “Science cannot make any statements about the existence of something we can’t observe, may that be gods or other universes.” She also said ““Electrons” is not an observable” (which in context clearly seems to mean that electrons are not observable). Those two claims imply that science cannot make any statements about the existence of electrons. That seems crazy.
Now perhaps Hossenfelder is adopting a non-standard theory of language. Perhaps, that is, she’s saying that scientific statements that contain the word “electrons” don’t imply that electrons exist. Rather, they are simply “shorthand” for a set of equations that are useful for prediction and control. So while science makes a lot of “electron” statements, none of them are “about the existence of electrons.” That’s sensible enough; it’s a normal instrumentalist thing to say.
But if instrumentalism is correct, then what do we make of Hossenfelder’s criticism of the multiverse theory? If the instrumentalist’s theory of language is right, then the multiverse theory doesn’t make any statements about the existence of a multiverse. The multiverse theory is just a “shorthand” for a set of equations in quantum mechanics. Those equations work. So the multiverse doesn’t “exist,” but that doesn’t mean that the multiverse theory is wrong. I’m not sure if this is the right way to understand Hossenfelder’s attitudes towards electrons and the multiverse, but it’s the best that I can make sense of what’s going on. I think there’s something missing here.
The second issue is far stupider. I think the real bone of contention between Goff and Hossenfelder is on the difference between assuming something and inferring something. Hossenfelder claims that it is unscientific to assume the existence of something unobservable. But it is legitimate to infer the existence of something unobservable as the best explanation of something that we do observe. Hence she’s not being inconsistent with what she says about electrons. They are not observable, but we nonetheless may infer that they exist as the explanation of certain experimental effects. What we may not do is assume they exist. And this explains her first tweet that Goff took issue with: “You don’t need maths to see that assuming the existence of other universes is unscientific.”
Hossenfelder is working within a conceptual framework where what you assume is very different from what you infer on the basis of those assumptions. Goff is not careful about drawing this distinction. Although his tweet says that scientists “assume” we live in a multiverse, he’s talking about inference as well. If you read the article that he linked to in his original tweet, you’ll see that Goff is rebutting an argument for the existence of the multiverse. A better summary would have been to say “Many physicists have inferred that we live in a multiverse - but their basic maths may be wrong!”
Now, Goff isn’t using terminology in a crazy way. What’s making this whole issue tricky is that we’re dealing with non-deductive inferences. That is, we’re trying to infer conclusions that are not strictly entailed by the data that supports them. It’s possible that some other theory is correct. So when we engage in non-deductive inferences, there’s always a large (in many case, infinite) range of theories we could choose from. From that infinite range, we select the theory that is “best.” But it’s always highly controversial which theory is best, and there’s (arguably) no way to identify the best explanation of some set of data. So when someone makes an inference to the best explanation, this will always involve making some assumptions about how to identify the best explanation. And it’s very common among philosophers to criticize an inference to the best explanation in those terms. “X is our data, and Smith assumes that Y must be the case (on the basis of X). But Smith’s reasoning is flawed. Here’s why.” Inferences to the best explanation are a mix of strict derivation and reliance on background assumption. So Goff is following the somewhat common practice among philosophers of using the word “assumes” to mean “infers (in a way that relies on background assumptions).”
But for Hossenfelder, assumption and inference are just different things. If you assume something, you aren’t inferring it. Assumptions are prior to inference. So of course it’s silly to “assume” the existence of a multiverse. We shouldn’t “assume” the existence of anything! In science, we infer that things exist (by inference to the best explanation). That’s how we know that electrons exist. And so we don’t need to do any math in order to know that no one should assume the existence of the multiverse.
But of course no one thinks that the multiverse is an assumption. It’s an inference from the core data of quantum mechanics. It might be a bad inference. But then again, the inferences that led scientists to accept the electron theory might be bad inferences, too. These are both non-deductive inferences that rest on contentious background assumptions. So while Goff might have been sloppy in his use of terminology, he was absolutely right to press Hossenfelder on how her argument should apply to the existence of electrons. Electrons are “unobservable” but we accept that they exist because they’re the best explanation of our data. The multiverse is (according to some) exactly the same: unobservable, but we know it exists because it best explains our data. (What makes it the best explanation? That’s a matter of our background assumptions, in both the case of the electron theory and the multiverse theory.)
I think this clarifies things sufficiently. And I hope that puts the whole stupid debate behind us.
Oh my unobservable God! That is a great great article! I am a philosopher my brother a cinematographer we had this same silly conflict years ago.