When it comes to QT, SRT, and GRT, what exactly does it mean to say a theory is “true”? I understand “successful” is in terms of a theory’s accuracy in what it predicts. But as I understand it, a theory’s equations “successfully” describing how matter behaves doesn’t entail that it’s conceptual framework is a true representation of what reality is actually like.
So how can it be justified to say that the conceptual framework of the theory is most probably true to reality based solely on its success in producing accurate equations? In a subjective inference like this wouldn’t it be a lack of prudence to draw such a conclusion based on the weight of one piece of evidence unless there’s a good reason to place so much weight on that evidence?
And if the claim is that there is a good reason, what is that reason that justifies attributing so much weight to that one piece of evidence? And that brings things back to the question I asked in my last post, which is, couldn’t empirical equivalence be used as an argument against attributing so much weight to that evidence?
I don’t really understand how your questions here relate to my last post.
Of course saying this presupposes some conception of a theory where it makes sense to say that it is true; i.e., it claims reality is a certain way. What might normally be called different “interpretations” (e.g. pilot-wave vs. many-worlds) are making different claims about reality and so are different theories in this sense.
Of course it doesn’t, and the problem of underdetermination just demonstrates this. But the scientific realist would say that the theory’s being true is a good explanation of its predictive success (this is an abductive, not deductive, argument); the epistemic structural realist would say more precisely that the theory captures something about the structure of reality while acknowledging that it may not be the whole story, ontologically or metaphysically speaking.
Emphasis added, note that I never claimed this. In fact, I directly claimed that theories ought to be judged on more than just empirical adequacy:
Maybe I wasn’t careful enough in my wording in the post I made that you were responding to. When I said,
By fields I meant in a more general sense that would include GRT as well. Does that help to clarify what my question that you were responding to was aiming at?
OK. But I find theory is used in an ambiguous manner where sometimes it means the conceptual framework as well as the equations that flow from it. You seem to be using it in the sense of just the conceptual framework by itself. Is that correct?
Or by “what reality is actually like” are you also referring to the equations and the predictions they make that flow from the framework which can be observed and therefore are empirically confirmed, as well as the idea that the conceptual framework is “what reality is actually like” that is not able to be empirically confirmed?
That leads into a question I have about underdeterminism. Here’s a quote from the SEP article:
All of these theories make all and only the same empirical predictions, so no evidence will ever permit us to decide between them on empirical grounds.
If empirical grounds means confirmation by some kind of observation, this seems to be referring to only the equations and their predictions associated with a theory and not at all the idea that the conceptual framework represents what reality is actually like since the implication of that statement is that we can decide on empirical grounds for theories that don’t have competing empirically equivalent theories.
In other words with equations and predictions in GRT vs Newton we can decide on empirical grounds which is more accurate based on the observation of the predictions whereas with empirically equivalent theories we cannot.
Right, but how does that apply to empirically equivalent theories? If that is the case, that seems to suggest more than one reality would have to exist, doesn’t it?
Does empirical adequacy have to do with confirmation by observation, or abductive inferences based on empirical evidence?
No. Mostly, I didn’t understand why you were asking questions that seemed to presuppose something I had directly contradicted (namely, that theories should be only be assessed based on their empirical predictions and nothing else).
I don’t understand what you are asking, exactly. You asked “what does it mean for a theory to be true”? To rephrase my answer (almost to the point of tautology), a theory is true if it makes correct claims about reality. Ambiguity of “theory” isn’t really an issue here; this applies rather broadly however you care to construe “theory”.
Now I really don’t understand what you’re asking. “What reality is actually like” is what reality is actually like. Conceptual frameworks, equations, predictions, etc describe reality. Some aspects of those descriptions, but not all, can be checked empirically. A theory is true if its descriptions of reality are correct. Forgive me for being confused at your questions, but this all seems rather obvious to me.
Again, you seem to just be rehashing the problem of underdetermination here. The problem isn’t just that, in cases of empirically equivalent theories, we can’t decide between them on empirical grounds. There is a broader problem in that the cases of empirically equivalent theories demonstrate that underdetermination is possible - given any theory, even if we don’t have an empirically equivalent alternative for it, there could be one that we don’t know about. (In fact this is a difficulty for all abductive reasoning / inference to the best explanation.) There are various responses to this problem, and I encourage you to look into them further; the scientific realism vs. anti-realism debate is the main context where you will find said responses.
By way of example, an epistemic structural realist might agree with you, in effect, that when we accept a scientific theory as an accurate description of reality we are primarily accepting that the equations of that theory reflect some aspect of the structure or relationships between entities in reality, while remaining more agnostic about the complete nature of those entities or of whatever is ontologically or metaphysically underlying the theory. (But there is a lot of room for nuance and different ways of spelling that out.)
It does not. Why would you think that?
A theory is empirically adequate if it makes correct predictions. So, it has to do with confirmation by observation, if you want to put it that way.
Let me go back to this. I would agree that the predictive success is evidence to support the inference that the conceptual framework is a representation of a true picture of reality, but I wouldn’t agree that it’s necessarily a good explanation of what reality is really like. That would need to be examined along with other evidence, not decided solely upon that one piece of evidence.
I have looked into them, although I’m sure there’s things about them, like with underdeterminism, that I may have missed. I did suspect that underdeterminism was referring to all theories. If that’s the case I think it boils down to the question of whether or not abductive reasoning can provide a reasonable amount of certainty if there is sufficient evidence, empirically or otherwise. At the root of denying that, in my opinion, is logical positivism.
Stepping back a bit here’s what I see. Theory as it’s used in this context includes both description and explanation. And though these terms can be used interchangeably, in this context here’s the distinction that I feel is not being recognized. Description is basically about prediction confirmed by observation, whereas explanation is about inference warranted by evidence. Or is there something I’m missing?
I would think that it is trivially obvious that a theory must be consistent with evidence applicable to the domain of that theory, or otherwise just serve as a placeholder. What is your definition of “evidence”?
I’m not really sure what you mean here. My claim was that the (potential) fact “Theory A is true” (which could be refined along structural realist lines, e.g. to “Theory A correctly captures the structure of some aspect of reality” or something like that) is a good explanation of the fact “Theory A makes successful predictions”. This can then be part of an abductive inference towards accepting “Theory A is true (or correctly captures the structure of some aspect of reality)”.
Now, depending on just how it is construed (e.g. it could be construed in a very strong realist way, or in a less metaphysically heavyweight way if you take an epistemic structural realist stance), Theory A may or may not provide a very deep explanation of why reality is the way Theory A describes. If you are objecting because you think that accepting Theory A as above always means accepting it as providing some deep explanation of why things are that way - I don’t think that’s the case, and you’re objecting against something I don’t see anyone arguing for. Otherwise though, again, I’m not sure what you mean.
Description is giving an account of what reality is like; explanation is giving an account of why it is that way. Absolutely, there’s a distinction here. But again, these activities are connected; the explanatory relations that hold in reality are at least in some way part of what reality is like, and so descriptions of what is going on often use terms that implicate those relations. Going back to something I said earlier:
All that being said, it really looks to me like you are (perhaps without realizing it) taking something like an epistemic structural realist stance with the questions you are asking here. If you’re open to book suggestions, I believe you’d find Ed Feser’s Aristotle’s Revenge of interest - in particular, the 3rd chapter discusses epistemic structural realism, and the 4th and 5th are relevant to the philosophy of relativity and quantum theory.
I’m not sure I follow exactly what you’re saying here. The inference under consideration is whether or not the implication of a conceptual framework is true about an unobservable aspect of reality, right? If not, what exactly is the inference being made?
If so, I think we agree here. We’re just expressing the same thing in different ways. But my point is focused on the fact that it is only “part” of an abductive inference. It isn’t the deciding factor. As I understand it, abductive inferences are usually based on multiple pieces of supporting evidence, not just one.
All I’m saying is that, no matter how good a possible explanation may seem, in order to decide how good it actually is it needs to be examined against all of the available relevant evidence. Until that happens it’s just an assumption, is it not?
I’m not sure that captures the distinction that’s under discussion. In the context of what we’re discussing, I would say that description is focused on observed effects and is mathematical in nature and addresses the questions of how matter behaves. Whereas explanation focuses on unobserved existence and is linguistic in nature and addresses the questions of what unobserved reality is like.
I’m not clear on what you mean by saying “explaining those observations”. As I understand it the observations aren’t what’s being explained. The question isn’t about what something that is seen is like. The question is about what something that isn’t seen is like.
That has to be inferred through abductive reasoning. Sure the relevant observations are used as evidence to argue for or against the inference, but they are not the focus of the question being asked. Is that more a less what you were getting at with that statement?
No, the inference being made is exactly what I said it was: whether some scientific theory or other is true (or correctly captures the structure of some aspect of reality), i.e. whether it makes correct claims about reality.
No one here has said we are inferring the truth of some scientific theory based on only one piece of evidence. You are arguing against an entirely imaginary opponent when you keep restating this, and misunderstanding what I am saying in the meantime.
My reference to this being part of an abductive inference is simply because an abductive inference requires more than just a good explanation of some evidence: it also requires comparing that explanation to other candidates (this is inference to the best explanation, not just inference to any good enough explanation), and at least considering whether there may be even better explanations that haven’t been thought of yet, for example. Of course part of assessing whether some explanation is good or not is seeing how it comports with all the available evidence. That has been assumed throughout this entire discussion.
Yes, the observations are what is being explained. Description answers “what”. Explanation answers “why”. The question is about why the things we see are the way they are.
These are completely idiosyncratic definitions of description and explanation on your part, Jim. They go a long way to explaining how often these discussions end up with you and your interlocutor talking past each other. That’s not what everyone else means by those terms - at least if I have any understanding of them!
I realize that you started off this thread by saying:
But it’s hard to even realize that you mean something so out of alignment with what everyone else means by relatively common words. So, with this source of confusion brought to light…
I hope at least that my statement of what I mean by those words clears up some of what I’ve said for you. To be honest, I’m still not sure what to make of everything you’ve said.
I’m a bit puzzled. It seems to me that something is a bit backwards here. I think it would be good to straighten out this issue before commenting on the rest of your post. To my thinking the first order of business needs to be answering the question of what the aspect of unobserved reality in question is like before answering the question about whether the conceptual framework of a theory is true in regards to that same aspect of unobserved reality.
We can establish whether a theory is successful, and if so we can use that as evidence for an inference that the conceptual framework of that theory supports. But it seems to me to establish the degree of certainty for the truth of the conceptual framework of a theory as reality, we first have to establish the degree of certainty that the inference that the conceptual framework supports is on the side of the weightier of the overall evidence.
Once that’s done then we can say if that conceptual framework falls into what seems to be the inference that is most supported by the evidence. And that framework may or may not be the one on the side of the weightier evidence. In other words, the success of the theory can support an inference about what unobserved reality is like, but it doesn’t determine that that inference will be what is most supported by the overall evidence. How does it make sense to just use one piece of a puzzle to come to a conclusion about what the entire puzzle is picturing without first trying to put all the pieces together?
Let me get this straight - you think we need to know what unobservable reality is like before we can make an abductive inference for the truth of some scientific theory about unobservable reality? How exactly are we supposed to do that without formulating a scientific theory based on what we can observe and making a reasonable inference regarding its truth? What other source of knowledge about unobservable reality are you adverting to?
This question should not be read as a denial that, e.g. philosophical/metaphysical reasoning can also lead us to knowledge about reality (including unobservable reality), but - or at least this is the way it seems to me - the nature of metaphysical reasoning is such that it gives us knowledge of the structure of reality at a very general level, compatible with all possible ways reality could be. For example, metaphysical reasoning may tell us that physical things are composites of form and matter, or that they are substances characterized by accidents, or something like that. But to find out what the specific nature of the physical world actually is, particularly at the unobservable level, we have to rely on scientific theories.
If that is all you mean - that our theories should be in accord with philosophical or metaphysical reasoning - then I already said that (albeit implicitly) a while back:
By coherence I meant not just internal coherence, but also coherence with all the evidence and with other bodies of knowledge (including philosophy). In retrospect, I should have made that more clear; my apologies.
Once again:
I also find it weird that you are characterizing the success of some scientific theory as “one piece of evidence”; scientific theories like quantum mechanics and relativity are supported by mountains of evidence.
I’m not saying you can’t do that. What I’m saying is that doing so doesn’t determine that the inference drawn from it is going to be the most probable of possible inferences when looking at all the evidence.
From the fact that a certain conceptual framework seems to provide the most accurate predictions it seems reasonable to infer that it’s so because that’s what unobserved reality is actually like.
But my point is that the way to determine the probability of whether that’s true or not requires not just looking at that one fact, but looking at all the other evidence as well in order to assess the probability of the truth of that inference.
You mean we have to rely to some degree or another on scientific theories, right?
From which of the observations in question is it possible to infer something about unobserved reality? The success of the theory is supported, or maybe better to say inductively confirmed, by observations. But it’s the success itself of the conceptual framework that is evidence for an abductive inference about unobserved reality.
The observations in question themselves are for the most part, if not completely, only relevant because of their inductively confirming the success of the conceptual framework of a theory as producing equations with the most accurate predictions.
If they were inductively confirming the truth of the conceptual framework as pertaining to unobserved reality they would be relevant evidence for that inference. But as far as I can tell, that’s just not the case. If they were, I don’t see how there would be much, if any, controversy over the truth of what they were confirming.
And as I’ve tried to explain multiple times now, I’m not talking about making an inference using only part of the evidence. I’m talking about about making an inference using all of the evidence. The theory may be only intended to (directly) explain a part of the evidence rather than all, but its coherence with all the evidence must be assessed as part of the inference for its being true (or correctly capturing some aspect of nature’s structure).
What you are saying here makes no sense to me. No, we are not inferring the success of the theory. The success of the theory is constituted by its making accurate predictions - we don’t see that the theory makes accurate predictions and infer “oh, the theory is probably successful”. We see that the theory makes accurate predictions and immediately know that the theory is successful, because that’s what it means for the theory to be successful.
What do you mean by this (the last sentence, in particular)? It seems to me that there’s controversy because this is a complex subject, and it isn’t easy to separate out all the competing theories / interpretations to compare them with each other and assess them all on the total evidence in order to do the abductive inference - not because such an inference just isn’t possible in principle, which is what you seem to be arguing.
This I’m not clear on. In the context of what we’re discussing as far as I can tell the inference being made by the theory is based only on the success of the conceptual framework of the theory. Am I missing something?
So if knowledge that is arrived at through induction isn’t an inference, what is it? And you mean we know tentatively, right? After all, at least to my understanding and depending on ones definition of induction, it is an inductive inference based on all the current observations confirming the success of the theory. How do you see it?
What I mean is, as you mentioned, we immediately know, in the tentative sense, that the theory is successful because we can see that it is from the observations. If we could see from the observations that the conceptual framework were true, we would immediately know that it was. But we don’t see that because what we’re talking about is unobservable reality.
Okay, first, let’s start with a comment on your use of the phrase “conceptual framework of the theory”. I haven’t bothered to engage with you on this point yet, but I should probably say that this is peculiar language that seems like extra verbiage to me. Earlier in this discussion I noted that I was using “theory” in fairly broad way, referring to the claims it makes about reality:
Using theory in this broad sense, as I have been doing, makes “conceptual framework of the theory” redundant. A theory includes its conceptual framework; there isn’t really any gap between them. (Though, to be sure, we can make a distinction between the equations of the theory and the concepts or interpretation that the theory applies to them.) I’m not sure if that is causing any confusion between us, but I might as well point it out just in case.
Second, to clarify what is being inferred from the success of a theory: someone taking a strong realist attitude towards a theory will infer that the theory is true, i.e. that it makes correct claims about reality (both in its equations and in its conceptual framework). Of course, this position is open to attack because of the problem of underdetermination: there could be empirically equivalent theories with the same level of success as the theory in question, but with a different interpretation of the equations, different ontological commitments, different conceptions of space and time, etc. (They may even have different equations, but more typically at least some of the equations will be shared, or the equations of one will be derivable from the equations of the other - this is the case for orthodox QM vs. pilot-wave theory, for example.)
A more moderate realist position, such as epistemic structural realism, would be less vulnerable to this attack. Someone taking an epistemic structural realist attitude towards a successful theory will infer that the theory correctly captures the structure of some aspect of reality. Very roughly, we could say that the epistemic structural realist will take the equations of the theory to be correct, but will be more open to looking at empirically equivalent theories (you might say, different conceptual frameworks) and relying on philosophical reasoning to judge which one is most plausibly correct.
Either stance is making an inference from the success of the theory, and the inference is grounded in the fact that the theory’s being true (or, for the epistemic structural realist, correctly capturing the structure of some aspect of reality) is a good explanation of why the theory is successful.
With all of that out of the way, I’ll try to clarify what I meant when I said:
by way of simple example. General Relativity is intended to explain the behavior of large-scale objects at high velocity under gravitational interactions. It is not intended to explain the behavior of very small-scale phenomena where we start to see quantum effects dominate. Nevertheless, to (ultimately) be an acceptable theory it must at least be coherent with the evidence that we have for the behavior of small-scale phenomena - and in fact, it isn’t (or at least, there are significant tensions preventing their full reconciliation), and this is how we know GR is not completely correct and why we are looking for a better theory to unite quantum effects with gravity.
Induction is an inference, but we don’t know that a theory is successful through induction - we know it by seeing that it makes correct predictions (i.e, after we have tested them). An inductive inference would be to infer that the theory will continue to make correct predictions, say, if we were to test it again tomorrow, or if we were to travel to Andromeda and test it there, or if we were to experiment on a newly discovered particle.
Obviously - that’s why we are talking about making an inference from the success of the theory to some claim about the nature of unobservable reality (i.e. that what the theory says about it is true, or that it in some way instantiates a certain structure).
I think it probably is what’s causing the confusion. I didn’t get that’s what you meant till now. But I’m not contesting that claims the equations are making about reality are true in the sense of their accuracy. And I’m not saying that a theory has to be both the most accurate and the best explanation regarding unobserved reality to be successful. I take successful to mean the most accurate predictions, no more no less.
But my contention is that just because a theory is true in that it makes successful predictions doesn’t determine that it is true in its representation of unobserved reality. And conversely just because it isn’t true in its representation of unobserved reality doesn’t mean it isn’t true in that its predictions are successful, unless successful is defined in a way that requires both to be the case. But the way I’m using success is only that it makes the most accurate predictions.
This seems to be the disconnect. My focus is on the claims being made about unobserved reality, not about the predictions. And I don’t see why they both have to go hand in hand. In my mind the theory can still be successful, i.e., make the most accurate predictions, without its claims about unobserved reality being true.
To tie them both together, in my opinion, is unjustified because they are ultimately addressing two totally different questions (although we might disagree on what exactly those questions are). And empirical equivalence, I think, further strengthens the argument against doing so by demonstrating that it’s possible for one of the answers to be right while the other is wrong.
If “that the theory’s being true” is only about its answer to the question about what reality is like, then I think we agree. Let me put it another way to see if we’re on the same page.
What I would say is the inference that the conceptual framework truly represents reality to some degree or another is grounded on the fact that the theory makes the most accurate predictions. And the idea is that the inference seems to make sense of why the predictions of the equations that flow from the conceptual framework of the theory are consistently empirically confirmed.
In other words, it seems like a reasonable inference that the conceptual framework, to some degree or another, is a true representation of unobserved reality based on the fact that the equations that flow from it are found to be the most accurate.
OK. I see what you’re saying. But I’m not quite sure how this applies to the issue being addressed.
OK. Think I got it.
So here’s what I think is going on in general with this discussion. It seems to me that it boils down to how to answer these two questions in particular:
What is the actual evidence in terms of what reality is like, i.e., is it just the success of the predictions of the theory, or each individual observation confirming the predictions.
And what does the accuracy of the predictions of a theory mean in terms of the question of what unobserved reality is like, i.e., does it mean that if the predictions are the most accurate, that its answer to the question of unobserved reality has to be true as well in order for the theory to be accepted?
Or does it mean that the accuracy of the predictions is only an indication that what the conceptual framework seems to be saying about what unobserved reality could be like is true, and the theory can still be accepted as partially true, i.e. that it gives the most accurate predictions, even if it were to be wrong about the answer to the question about unobserved reality?
Would you agree that those are the issues that we should focus on in order for the conversation to move forward?
Yes, that is the sense I have been using it in as well.
Again, I agree. The problem of underdetermination and the existence of empirically equivalent theories makes this clear. That’s why an abductive inference to the truth of the theory’s claims about unobservable reality requires philosophical evaluation of the theory, and comparison to other theories in terms of coherence (both internal and external), parsimony, explanatory power, etc. But unless you think the possibility that a theory is true has nothing at all to do with why it might be successful - bit of a stretch, in my opinion - it’s hard to see why such an inference could not be justifiable in principle.
Which I guess you actually do agree with, if I’m reading you correctly here?
To be honest I’m not sure we are not just saying all the same things in different ways, at this point.
I don’t see a difference there. Or maybe more precisely, perhaps you could make some kind of distinction, but then it is clear the latter that matters - the individual observations are the actual empirical data that the theory is supposed to explain; the inference to the best explanation is really about explaining the data and only secondarily about explaining the success of the theory, even though I have been speaking more about explaining the success of the theory. But the two are intertwined - the more successful the theory, the more data it encompasses.
Again, certainly the latter - as the existence of empirically equivalent theories shows, a theory can be false in some of its claims about reality while still making accurate predictions. I think we’re on the same page here.
You might want to note that after you pointed out (correctly) that “conceptual framework” was unnecessary verbiage in the context of the discussion, @Jim just went and used it again.
OK. From what I can gather, it seems like I’m tracking with you for the most part. I guess where I’m not real clear on what’s going on is in regards to the evidence.
I’m assuming the data being discussed here is the observations confirming the predictions, right? If so, could you elaborate a bit more on what exactly you mean by how the theory is explaining the data?