Yes, I get that, and like I said in my last post, "That falsification only gives probabilistic support to the contrary doesn't change that seeing something a theory predicts doesn't even make it more likely, never mind certain, that the theory is true."

Seeing something to the contrary of what the negation of your theory predicts does show your theory true, though. And if the negation of your theory only has something predicted to be unlikely, then seeing that something only makes your theory likely true. I even mentioned this in the OP, and have quoted it again later in the thread since:

But this does not imply that all beliefs not yet shown false are equal. Beliefs not yet shown false can still be more or less probable than others, as calculated by methods such as Bayes' theorem. Falsification itself can be considered just an extreme case of showing a belief to have zero probability: if you are frequently observing phenomena that your belief says should be improbable, then that suggests your belief is epistemically improbable (i.e. likely false), and if you ever observe something that your belief says should be impossible, then your belief is epistemically impossible (i.e. certainly false). — Pfhorrest

Your "if Q then P" is, as you said, a equivalent to "if not-P then not-Q". If that is (probably) true, and Q is true, then P is (probably) true. But that's the falsificationist method, not the confirmationist method.

The confirmationist method would say that if "if P then Q" is (probably) true, and Q is true, then P is (probably) true, and that's simply not a valid way of reasoning, even with the "(probably)"s in there.

This is contrary to the idea that X ought to be tentatively accepted until falsified — Kenosha Kid

I never said it ought to be, only that it may be. Pending evidence either way, both X and ~X are permissible beliefs. To say that pending evidence either way, both X and ~X are impermissible beliefs (what I mean by "cynicism") would make it impossible to ever have evidence either way (because you would need some beliefs to be the evidence, but you couldn't hold those without others that you also aren't permitted to hold yet, ad infinitum), and so impossible for any belief in anything to ever be permissible.

(It might be worth reiterating here that by "belief" I don't mean anything anywhere near dogmatic, merely "thinking something is true". There are all kinds of things we think are true, but are perfectly open to evidence that they're not).

other ideas one finds uncontroversial — Kenosha Kid

Is it okay to believe in those things you think are uncontroversial without first proving that they're correct from the ground up? My "liberalism" says yes, and its negation I call "cynicism" says no.

Your "if Q then P" is, as you said, a equivalent to "if not-P then not-Q". If that is (probably) true, and Q is true, then P is (probably) true. But that's the falsificationist method, not the confirmationist method.

The confirmationist method would say that if "if P then Q" is (probably) true, and Q is true, then P is (probably) true, and that's simply not a valid way of reasoning, even with the "(probably)"s in there. — Pfhorrest

This is not correct, you keep presenting it backwards; which amounts to refuting a strawman. Sticking with the present example, the confirmationist method says precisely that tool marks would reasonably be thought, based on an enormous body of experience, to be the main sign of artificiality. The presence of tool marks, for all intents and purposes, confirms artificiality, but can never prove it.

The problem with simply saying that falsification "does the heavy lifting" is that it ignores that fact that nothing can be falsified without other things being confirmed; or to put it in other words anything is falsified only to the degree that other things are confirmed.

This is contrary to the idea that X ought to be tentatively accepted until falsified
— Kenosha Kid

I never said it ought to be, only that it may be. — Pfhorrest

all beliefs should be considered justified enough by default to be tentatively held (the liberal part) until reasons can be found to reject them (the critcal part) — Pfhorrest

So it is the liberal part I'm referring to: that any given belief should be considered justified enough to be tentatively accepted. This would include any absurd yet so far untested belief that I might make up: that spiders are telepathic, or that The Great Geoff lives on an asteroid orbiting the black hole at the centre of our galaxy, or that the CIA are controlled by a secretive Inuit conglomerate.

The things I believe are not quite this random I hope. I believe the things I do not just because they have not yet been falsified, but also because there is some reason to do so. Popper's criterion is not just that an idea isn't falsified, but that it is nonetheless falsifiable, i.e. we can test it, and see if the world fails to work as if the idea is false. It is not proof because another test might falsify the idea yet, but it is a stronger grounds to believe than 'hasn't been falsified'.

Pending evidence either way, both X and ~X are permissible beliefs. To say that pending evidence either way, both X and ~X are impermissible beliefs (what I mean by "cynicism") would make it impossible to ever have evidence either way (because you would need some beliefs to be the evidence, but you couldn't hold those without others that you also aren't permitted to hold yet, ad infinitum), and so impossible for any belief in anything to ever be permissible. — Pfhorrest

But this is not what I suggested. I said that I can suspend judgment on either given no facts to support either. We aren't obliged to take a firm position on everything. Do I believe Jesus lived or not? Neither. I don't know, and I don't really care. It's a matter of supreme indifference to me.

Is it okay to believe in those things you think are uncontroversial without first proving that they're correct from the ground up? — Pfhorrest

Yes, but then I don't subscribe to the view that ideas either need to be proven or distrusted any more than I subscribe to the view that ideas must be falsified or held tentatively as true. They are false dichotomies in my view.

Pragmatically, beliefs are tools for predicting the world, the best founded beliefs being those best aligned with experience, the worst founded being those conflicting with evidence (falsification). Those exactly in the middle which are neither falsified nor supported (as opposed to proven) are likely useless, probably meaningless too. The more evidence for an unfalsified idea, the stronger the basis for belief.

However strongly justified a belief, a reasonable person must reject it the moment they see it falsified. In the meantime, so long as the belief is both falsifiable and consistent with the world, the believer is perfectly justified in holding it to be as if it were true, i.e. to have assumptions about the world.

This is not correct, you keep presenting it backwards; which amounts to refuting a strawman. — Janus

I’m refuting the thing that’s called “confirmationism” in distinction from “falsificationism”. If that’s not the thing that you support, then that’s great, but that’s the thing that’s called that name in philosophy of science. I’ve been saying since the beginning that what you’re saying isn’t counter to what I’m saying, because what I’m say is precisely an argument against that form of inference. It’s not a straw man just because you don’t support it; I didn’t start out arguing against you, saying this is what you believe, you just came in arguing nominally against what I was saying, with things that weren’t actually against it.

Fair enough; you're pointing out the problem with the deductively invalid form of argument "If P then Q, Q therefore P", and arguing against verification because you associate that invalid form of argument with verificationism or confirmationism.

I'm doubtful that verificationists (for example Ayer or the Logical Positivists) were stupid enough to believe that argument is valid, so I think they must have been talking about something else.

The something else I think they were talking about is that repeated observations coupled with an enormous accumulated body of theory based on those observations does give us good reason to think in many contexts that when we observe what is predicted we are warranted to hold (always provisionally) many things to be confirmed.

In short I don't think there was ever any coherent ""confirmationism" in distinction from "falsificationism"": or vice versa. If you still disagree that's fine; we'll just have to agree to disagree.

you're arguing against the deductively invalid form of arguemnt "If P then Q, Q therefore P", because you associate it with verificationism or confirmationism — Janus

Just confirmationism, not verificationism as in the Positivists. Confirmationism is something broader (as in less specific, less comprehensive) than verificationism; non-verificationists can still be confirmationists.

And as I've just said a bunch of times, it's not just about deductive validity. It's that that form of argument doesn't even give probablistic support to P. If you reverse (or equivalently negate) the antecedent and consequent of the first premise, then it does, yes, but that's precisely the switch from confirmation to falsification.

repeated observations coupled with an enormous accumulated body of theory based on those observations does give us good reason to think in many contexts that when we observe what is predicted we are warranted to hold (always provisionally) many things to be confirmed. — Janus

That's precisely the same thing as "it's probable that if P [the body of theory] then Q [the predicted observations], Q, therefore probably P". Which there's no good reason to think, and plenty of good reason to think otherwise. (I'm avoiding the word "valid" here because you'll think I'm talking about deduction again).

If you reverse the P and the Q, or equivalently negate them, then there is good reason to think that way, yes. But that's precisely the switch from confirmation to falsification.

So it is the liberal part I'm referring to: that any given belief should be considered justified enough to be tentatively accepted — Kenosha Kid

It's the "enough" part that matters there. It's not that you should tentatively accept it, but that you should not demand proof from others if they want to tentatively accept it. (Nor, if you yourself feel inclined to accept it, demand proof from yourself or else reject it; if it seems true to you, go ahead and believe it).

This would include any absurd yet so far untested belief that I might make up: that spiders are telepathic, or that The Great Geoff lives on an asteroid orbiting the black hole at the centre of our galaxy, or that the CIA are controlled by a secretive Inuit conglomerate. — Kenosha Kid

Yes, if you're inclined to believe those things, then go right ahead, and if someone else is, let them, unless you have reason to suggest you or they should not. (NB that this doesn't mean that you have to accept whatever nonsense someone else is inclined to believe, just that if either of you wants to change the other's mind, you need to show that they're wrong, not just point out that they can't show that they're right).

But this is not what I suggested. I said that I can suspend judgment on either given no facts to support either. We aren't obliged to take a firm position on everything. Do I believe Jesus lived or not? Neither. I don't know, and I don't really care. It's a matter of supreme indifference to me. — Kenosha Kid

Then that is not the thing I call "cynicism", and I think that's perfectly fine.

the view that ideas must be falsified or held tentatively as true — Kenosha Kid

That's not my view. My view is that non-falsified beliefs are permissible beliefs (that you can hold them as true without committing an epistemic error), not that they are obligatory beliefs (that you must hold them as true or else you're committing an epistemic error).

The more evidence for an unfalsified idea — Kenosha Kid

The inability to ever have evidence for something, rather than merely against the alternatives, is the whole point of falsificationism.

However strongly justified a belief, a reasonable person must reject it the moment they see it falsified. In the meantime, so long as the belief is both falsifiable and consistent with the world, the believer is perfectly justified in holding it to be as if it were true, i.e. to have assumptions about the world. — Kenosha Kid

:up: That's exactly my position as well.

(Nor, if you yourself feel inclined to accept it, demand proof from yourself or else reject it; if it seems true to you, go ahead and believe it). — Pfhorrest

Something 'seeming true' is a reason to believe it, not the believing of it. Some of those are rational, such as empiricism; some are not (e.g. I cannot stand the idea that... I don't want to live in a world that...). What I'm getting at is that these are not equal. I actually wouldn't advise that someone goes ahead and believes something true that is not falsified or evidenced, especially if it's opposite or negation is evidenced albeit unproven, but that aside simply down to the lack of good reason to believe.

The inability to ever have evidence for something, rather than merely against the alternatives, is the whole point of falsificationism. — Pfhorrest

Not at all, you can always have evidence for something. A witness testimony is evidence that the accused was at the scene of the crime, for instance. It just isn't proof. Evidence for something is always incomplete; evidence against it is always terminal. That is the whole point of falsification as I understand it.

We're not a million miles apart but the above distinction is the difference. Evidence is not all or nothing. There are many degrees between a completely arbitrary unfalsified belief and a well-founded unfalsified belief.

However strongly justified a belief, a reasonable person must reject it the moment they see it falsified. In the meantime, so long as the belief is both falsifiable and consistent with the world, the believer is perfectly justified in holding it to be as if it were true, i.e. to have assumptions about the world. — Kenosha Kid


:up: That's exactly my position as well. — Pfhorrest

To be "consistent with the world" just is to be confirmed by observation, which is of course the same as to not have been (yet) falsified. The belief that tool marks are the main sign of artificial rock formations does not derive from falsifying anything but from countless confirmations that those rock formations displaying tool marks are indeed artificial. So, that is at odds with your "liberal" notion that we do and should, "believe whatever we like as long as it has not been falsified". This point has been the crux of all my responses to you in this thread.

Yes, what you say here is essentially what I (and others) have also been arguing against the OP.

especially if it's opposite or negation is evidenced albeit unproven — Kenosha Kid

I consider that a good (but not complete) reason to disbelieve something. You probably shouldn’t believe things that are probably false. I’ve said above (in the OP and requoted twice since then) that I’m not advocating a black and white view: things that are not completely falsified can still be epistemically unlikely, and you’re not completely wrong to believe those, but you’re taking a big risk. (I actually think this is perfect analogous to risky but not impermissible behavior, as well: it’s not wrong per se but you probably shouldn’t do it).

Not at all, you can always have evidence for something. A witness testimony is evidence that the accused was at the scene of the crime, for instance. It just isn't proof. Evidence for something is always incomplete; evidence against it is always terminal. That is the whole point of falsification as I understand it.

We're not a million miles apart but the above distinction is the difference. Evidence is not all or nothing. There are many degrees between a completely arbitrary unfalsified belief and a well-founded unfalsified belief. — Kenosha Kid

I would say instead that there are many degrees between a completely falsified belief and a mostly-unfalsified one. See the discussion above with Janus about the different kinds of probabilistic inferences. If P probably implies Q and Q seems to be the case, that doesn’t even give you the tiniest additional support for P; but if Q implies P (or equivalently not-P implies not-Q) and Q seems to be the case, then that does give incomplete support to P, but only because it’s equivalent to a probabilistic falsification: something that’s likely to be false if P were false seems to be true, so P is probably true, precisely because not-P is probably falsified.

To be "consistent with the world" just is to be confirmed by observation — Janus

It’s to be not falsified, but that’s not the same thing as being confirmed.

The belief that tool marks are the main sign of artificial rock formations does not derive from falsifying anything but from countless confirmations that those rock formations displaying tool marks are indeed artificial. — Janus

Where our beliefs originate from is not the issue at hand here. How to decide between conflicting beliefs is. You’re also changing which belief you’re talking about in the example. First you were talking about the belief that the face on Mars is artificial. Tool marks were the implication of that belief that we would check that belief against. Now you’re talking about the belief that that implication is true. We would need a different implication from that belief in the first implication in order to test whether the belief in the first implication were false or not, and that testing would have to be done falsificationistically.

Where our beliefs originate from is not the issue at hand here. — Pfhorrest

Of course it is; beliefs that derive from well-examined repeated experience should inspire more confidence than those which do not. I'm not changing the example at all. The belief that the face on Mars is artificial can be checked by examining whether or not tool marks are evident. If they were evident we would have good reason to believe that the face is artificial, if they were not evident we would have no reason to believe the face is artificial.

Of course it is; beliefs that derive from well-examined repeated experience should inspire more confidence than those which do not. — Janus

Which is to say, beliefs that have survived many potential falsifications.

I'm not changing the example at all. The belief that the face on Mars is artificial can be checked by examining whether or not tool marks are evident. If they were evident we would have good reason to believe that the face is artificial, if they were not evident we would have no reason to believe the face is artificial. — Janus

Let P = "the face on Mars is artificial"
Let Q = "there are tool marks on the face of Mars"

We're previously been discussing how we would test P. My entire point on that subject has always been that P implying Q, plus Q being true, does not give support to P; only not-P implying not-Q (or equivalently, Q implying P), plus Q being true, would give any support to P.

In your previous post before this one, you mentioned "The belief that tool marks are the main sign of artificial rock formations". That's not our "P". That's "if Q then P" (if I'm generous in interpreting you there, and you didn't mean "if P then Q" instead).

Let R = "if Q then P".

We believe R based on repeated observations wherein it is never the case that Q and not-P, which suggests to us that not(Q and not-P), which is equivalent to Q implying P. That's what it seems like you're saying, and that's a perfectly fine reason to come to believe that.

But now say we want to test that belief against other possibilities, maybe because someone else doesn't think the evidence suggests that belief, or just because we're undecided between multiple interpretations of the evidence ourselves.

To test R, we need to derive an implication from not-R. Call that implication not-S. I don't know what that implication would be off the top of my head, but maybe you can come up with something that would have to be false if "Q implies P" was false, and we can call that thing "S".

If R merely implied S, and S were true... that wouldn't give us any support for R. We'd still be free to believe R, just because it seems true to us, but if it seemed false to someone else, or we were undecided on it ourselves, we wouldn't have anything decide that disagreement or indecision with... unless the non-R possibility also implied a non-S observation, in which case we could rule out that possibility, but still not all possible alternatives to R.

Only if not-R (any and all alternatives to R, any scenario where R was not the case) implied not-S, and S were true, would that give us support for R.

I would say instead that there are many degrees between a completely falsified belief and a mostly-unfalsified one. — Pfhorrest

Both can be true.

The characteristic decay chains of the Higgs boson in the LHC data are evidence, not proof, that the Higgs exists, sufficiently so to earn Peter Higgs a Nobel prize. It is not evidence against ~Higgs, since there are potential theories that could explain the same data with more than one particle. But the more signals corresponding to expected decay chains we see (more have been discovered very recently), the better founded the belief that the Higgs mechanism is a good model of reality. It is not proven, but nor does it have the status of 'merely unfalsified' which might apply to something that has not been tested at all.

By "mostly unfalsified", I assume you mean falsified with less than 100% certainty. An example might be Trump's claim of election fraud, insofar as the few concrete claims have been mostly thrown out or pulled as they don't agree with fact. Nonetheless there is no obvious feasible means of completely killing off the broader claim.

Third, where evidence against not P is evidence for P. Is the ball under the left cup or the right? Assuming the ball is under one of the two cups, falsifying the theory it is under the left cup is identically evidence for it being under the right cup. There's no distinction between falsifying ~P and verifying P. (Of course, by sleight of hand or a tricksy table, it might be under neither, but the *belief* it is under the right cup is affirmed, though not confirmed.)

That's not our "P". That's "if Q then P — Pfhorrest

LOL, I've already explicitly stated that if P is the proposition that the face is artificial, and Q is the presence of tool marks, and if we understand tool-marks, based on extensive confirmatory experience, to be the main sign of artificiality, then the proper equation is "if Q, then P"; it was I who pointed that out to you much earlier on. I also pointed out that it is the hidden confirmatory premise in your falsificationist equation: "if P then Q, not-Q, therefore not-P".

Yet you continue to try to tendentiously distort everything to pass through your preferred lens of falsification. As I've said nothing you've presented has given me any reason to change my view that verification and falsification are the two faces of the one coin. I think to get clear on this you will continually need to remind yourself that there is no deductive certainty, on account of there always being unproven premises, in either verification or falsification.

It is not evidence against ~Higgs, since there are potential theories that could explain the same data with more than one particle. But the more signals corresponding to expected decay chains we see (more have been discovered very recently), the better founded the belief that the Higgs mechanism is a good model of reality. — Kenosha Kid

Why do those observations not equally lend support to the other theories that are just as consistent with them? I suspect you can actually provide an answer, because I trust that working physicists actually are smart enough to be using sound methods.

So I’m expecting that there is something about each of those alternative theories that is less consistent with all of the observations than the Higgs is; that each theory concords with fewer observations, or expects those observations with lower probability, or some combination thereof. In other words, the observations weigh against the other theories more than they do against Higgs. That is just a probabilistic version of falsification, which I mentioned in the OP and have requoted three times since then.

By "mostly unfalsified", I assume you mean falsified with less than 100% certainty. — Kenosha Kid

I mean a theory that at worst says that the observations we’ve made are a little unlikely. Complete falsification is when a theory says that the observations that we see are not possible, which thus renders that theory (epistemically) impossible (i.e. certainly false). A theory saying the observations we see are merely improbable in turn makes the theory (epistemically) improbable, which is a lessened version of falsification, or a partial falsification if you will.

Third, where evidence against not P is evidence for P. Is the ball under the left cup or the right? Assuming the ball is under one of the two cups, falsifying the theory it is under the left cup is identically evidence for it being under the right cup. There's no distinction between falsifying ~P and verifying P. — Kenosha Kid

Like I’ve been saying to Janus over and over, that’s beside the point. Sure, if you can show that not-P implies not-Q, and that Q, then you can show that P, via falsifying not-P. But that’s not what falsification was ever against.

What it’s against is saying that if you can show that P implies Q, and that Q, then you can show that P. That’s what confirmationism says, and what falsificationism denies.

Either of those can be rendered probabilistic instead of absolute as you like, and the same difference applies.

LOL — Janus

You know, I’m getting kind of tired of being condescended to by someone with such obvious reading comprehension difficulties.

I know you were talking about the same P and Q. But as I just said in my last post, we were talking before about how to test whether P or not using Q, and then you suddenly switched to talking about why we think that Q implies P.

I also know you were talking about Q implying P already (and how that’s equivalent to not-P implying not-Q), but that reversal from P implying Q just is switching from confirmationism to falsification.

"If P then Q, Q, therefore P" is confirmationism.
"If Q then P, Q, therefore P" is falsificationism, because it's equivalent to
"If not-P then not-Q, Q, therefore P".

You’re not showing that confirmation is hidden within falsification, you’re showing that falsification is the thing we need to do instead of confirmation, which is my whole point.

But I’ve explained all this many times before and you didn’t get it then so I don’t know why I expect you to get it now either.

Why do those observations not equally lend support to the other theories that are just as consistent with them? — Pfhorrest

That was precisely my point. Other than the absence of a historical competitor theory, the initial evidence is no more for one theory than another which yields those particular outcomes. As one increases the number of successful predictions, one ought to eliminate possible competitors, else the two theories are empirically indistinguishable. That process is ongoing, but there's no point at which LHC data will suddenly rule out an alternative model; in fact, I always assume that, in future, observation will lay waste to most of our models. The alternative is that I live in a privileged era.

Point being that we increase our faith in the model the more it fails to be falsified, without it ever being proven true. And this is not because we have falsified particular known competitors, and certainly not because we've falsified ~Higgs, but because we have narrowed down what a competitor theory can predict that is different to the Higgs model. One doesn't actually have to formulate the competitor theory to falsify it: it is sufficient to know that, as each data point is collected that is consistent with Higgs theory, so long as no data point is collected that rules it out, whatever potential competitor theories might be formulated are either falsified or equally consistent with the data do far, i.e. are *like* the Higgs theory to an increasing extent.

This gives us some confidence that, while Higgs might yet be ruled out (and probably will be), it is encouragingly close to reality. And it is this increasing confidence that increases our belief in Higgs. We did not start out with that level of confidence, or that strength of belief.

Like I’ve been saying to Janus over and over, that’s beside the point. Sure, if you can show that not-P implies not-Q, and that Q, then you can show that P, via falsifying not-P. But that’s not what falsification was ever against. — Pfhorrest

I included this partly for completeness and partly because you seem to interpret evidence for P and against ~P as purely falsificationist, i.e. only ~~P. But this is P so it's an erroneous distinction imo. In this class, there is no distinction between falsifying ~P and verifying P.

Point being that we increase our faith in the model the more it fails to be falsified, without it ever being proven true. And this is not because we have falsified particular known competitors, and certainly not because we've falsified ~Higgs, but because we have narrowed down what a competitor theory can predict that is different to the Higgs model. One doesn't actually have to formulate the competitor theory to falsify it: it is sufficient to know that, as each data point is collected that is consistent with Higgs theory, so long as no data point is collected that rules it out, whatever potential competitor theories might be formulated are either falsified or equally consistent with the data do far, i.e. are *like* the Higgs theory to an increasing extent. — Kenosha Kid

I already tried exactly this line of argument (beginning here and here). It won't work.

As far as I can tell, nothing will budge @Pfhorrest from his position. Nor should anything, in a sense, since the principles in play are not themselves falsifiable. That's irony or necessity, as you like, I suppose.

Personally, I think classical logic is just too primitive a tool here: material implication is not a good model for causation or for the relation between a theory and a prediction of that theory. Conditional probability is a better fit for both, and if you take a Bayesian approach you still get falsification as a special case, while picking up a reasonable treatment of confirmation.

*

But, hey, you do you, @Pfhorrest. It looks more and more like you're not getting the kind of feedback you wanted, and endlessly defending your fundamental principles has become tiresome for you, which is certainly understandable. (Some people are here precisely in order to have the same argument over and over again and prefer saying exactly the same thing over and over again, so we had no way of knowing you weren't one of those.) Is there an earlier post that makes it clearer what sort of feedback would be more useful to you?

Pretty much the whole process you describe is consistent with the approach I advocate, as I already said to Srap when he said similar things earlier. It’s the weeding out of alternatives (even if we haven’t enumerated them yet) that progresses our knowledge, and multiple models that equally well survive that process are not elevated by the process but are merely equal co-survivors of it.

As far as I can tell, nothing will budge Pfhorrest from his position. Nor should anything, in a sense, since the principles in play are not themselves falsifiable. That's irony or necessity, as you like, I suppose. — Srap Tasmaner

Not empirically falsifiable of course because these are not empirical issues, but open to logical falsification, like reductio ad absurdum or something, sure. I am not clinging to these principles against arguments to the contrary, because nobody has actually said anything to the contrary of my principles. They’ve only been attacking strawmen and saying things I already agree with as though that refutes the things I think, or conflating multiple things together so as to sneak in something unsupported along with something I already agree with.

Conditional probability is a better fit for both, and if you take a Bayesian approach you still get falsification as a special case — Srap Tasmaner

I said exactly that in the OP, and have referred back to or requoted it many times since.

Is there an earlier post that makes it clearer what sort of feedback would be more useful to you? — Srap Tasmaner

No, I’m not looking for anything in particular. I just wish that clarifying that I am not actually against the things you think you’re refuting me with would settle these nominal disagreements. The kinds of responses were fine at first (besides Isaac objecting to having the discussion at all), I just wish it wouldn’t go around in circles so much.

I said exactly that in the OP, and have referred back to or requoted it many times since. — Pfhorrest

You really didn't, and if you had said it you'd be wrong. Conditional probability is a whole different animal from material implication, and no adding of "probably" changes that, as David Lewis showed, like, forty years ago.

That's why I keep saying you have to choose between the logico-deductive model and the Bayesian model. You don't think you have to choose, but you're wrong.

I said exactly that in the OP, and have referred back to or requoted it many times since.
— Pfhorrest

You really didn't, and if you had said it you'd be wrong — Srap Tasmaner

If I said the thing that you just said, I'd be wrong?

What I said was:

Beliefs not yet shown false can still be more or less probable than others, as calculated by methods such as Bayes' theorem. Falsification itself can be considered just an extreme case of showing a belief to have zero probability — Pfhorrest

And you said:

if you take a Bayesian approach you still get falsification as a special case — Srap Tasmaner

Sounds like the same thing to me.

Conditional probability is a whole different animal from material implication, and no adding of "probably" changes that, as David Lewis showed, like, forty years ago. — Srap Tasmaner

Can you please elaborate on this? My adding of "probably" to the conditionals under discussion was not meant to be a formal thing at all, but a loose way of phrasing the idea that, as I said in the OP:

if you are frequently observing phenomena that your belief says should be improbable, then that suggests your belief is epistemically improbable (i.e. likely false), — Pfhorrest

Which as I understand it is what Bayes is all about.

Pretty much the whole process you describe is consistent with the approach I advocate, as I already said to Srap when he said similar things earlier. It’s the weeding out of alternatives (even if we haven’t enumerated them yet) that progresses our knowledge, and multiple models that equally well survive that process are not elevated by the process but are merely equal co-survivors of it. — Pfhorrest

Yes, we're not a million miles apart. Subtract your tentative holding true of untested beliefs, which the above does not require, and we're more or less on the same page.

Subtract your tentative holding true of untested beliefs, which the above does not require, and we're more or less on the same page. — Kenosha Kid

I also don't require it, I only permit it, so it sounds like we agree.

:up:

If your use of "probable" isn't formal, you're not going to be "calculating" anything.

Bayes' rule allows for your confidence, or your subjective degree of belief, to increase given new evidence. You allow change only in the sense that some alternatives are no longer viable, but the surviving theory still has exactly the same status it had before. If that's a Bayesian view of evidence, it's one I'm not familiar with.

Anyway, look into or don't. It's your project, not mine.

Can you please elaborate on this? My adding of "probably" to the conditionals under discussion was not meant to be a formal thing at all, but a loose way of phrasing the idea that, as I said in the OP: — Pfhorrest

Let x be a real number between 0 and 1
(1) P( x is rational ) = 0
(2) x is not rational.
Does (1) materially imply (2)? It does not, x=0.5 is a counter model. It can be true that x is rational even when P( x is rational ) = 0.

(1) clearly does support (2) in some way. But it's not a material conditional (1) -> (2) as there's a counter model. If (1)->(2) is false, then the material contrapositive not(2)->not(1) is false too as they're logically equivalent. Clearly observing not(2) is amazing evidence that not(1) ("They said it could never happen but it did!"), but it's not a raw modus tollens refutation - it's some different form of inference.

To put a super fine point on it: from not(2) it should be inferred somehow that not(1), but not(2) does not materially imply not(1).

You’re not showing that confirmation is hidden within falsification, you’re showing that falsification is the thing we need to do instead of confirmation, which is my whole point. — Pfhorrest

"If P then Q, Q, therefore P" is confirmationism.
"If Q then P, Q, therefore P" is falsificationism, because it's equivalent to
"If not-P then not-Q, Q, therefore P". — Pfhorrest

No this is precisely the point your position relies on which is incorrect. "If P then Q, Q, therefore P" is simply an invalid deduction. Confirmationism is an inductive, not an invalid deductive, thought process. As I pointed out already many times, but which you, due to your inability to countenance anything which is counter to what you have stipulated, the correct formulation for confirmationism is "if Q then P", where Q is believed to be, not a logically necessary sign of P, but a very strong inductive support for it.

So, in the example of the "Face", tools marks are thought to be a sign strongly suggestive of artificial structures, and that conjecture is confirmed, although not logically proven, by countless examples drawn from experience. If tool marks are observed then the proposition that the structure is artificial would be confirmed, which simply means that we have good reason, according to our experience, to think that the structure is artificial.

This has nothing to do with falsification other than that, if tool marks were not observed, then the proposition that the structure is artificial would be falsified, not proven false, mind, which means that we would not have good reason to believe it is artificial. Verifying and falsifying are thus two sides of the one coin; and you have provided no arguments, but instead merely the same stipulations over and over, to support your mere assertion that they are not.

It would make it seem much more like you are arguing in good faith, rather than doubling down on your position if you actually responded to what I am arguing here, (and others have argued) rather than playing a 'tit for tat' game of accusing me of poor reading comprehension, simply because I won't acquiesce to your stipulations.

It would make it seem much more like you are arguing in good faith, rather than doubling down on your position if you actually responded to what I am arguing here, (and others have argued) rather than playing a 'tit for tat' game of accusing me of poor reading comprehension, simply because I won't acquiesce to your stipulations. — Janus

I accuse you of poor reading comprehension because I have been responding to your arguments, and you never seem to understand the responses. You are mostly putting forth things that I don't disagree with, as though they disprove the things that I am saying. So I'm not going to put forth counter-arguments to show that the things you're putting forth are wrong, because they're mostly not. They're just beside the point of anything that I was saying in this thread, not against it, and I'm trying to show you why that is.

For example:

tools marks are thought to be a sign strongly suggestive of artificial structures, and that conjecture is confirmed, although not logically proven, by countless examples drawn from experience — Janus

This is what I mean about you changing the focus of the discussion. First we were talking about how we could test whether or not the Face was artificial. We could test that using the implication of tool marks to artificiality, and we were discussing the right way to use that implication to test for artificiality. But now you're talking about how we could know whether tool marks imply artificiality. That's a different thing. In a complete investigation that is a further question that we could step back and ask too, but it's not the same question we were asking about at the start.

Anyway, on to the meat of things.

"If P then Q, Q, therefore P" is simply an invalid deduction. Confirmationism is an inductive, not an invalid deductive, thought process. — Janus

I have repeated over and over again that I'm not simply talking about deductive vs inductive implications. I'm talking about the direction of implication. Inferrring from "if P then probably Q", and "probably Q", to "probably P", all merely probabilistically, is still not a good inference.

(I'm saying "good" instead of "valid" here so you don't think I mean non-probabilistic deduction; I've done that before as well. Also, probabilistic inference is not the same thing as induction, but let's leave that aside for now).

Meanwhile, inferring from "if Q then probably P, and "probably Q", to "probably P", even if it's all merely probabilistically, is a good inference.

And that is the whole point of falsificationism, because "if Q then P" is logically equivalent to "if not P then not Q", as you already know.

(This is another place where you've been just talking past me. You speak as though you pointed out the equivalence of those to me, and that that defeated my point, when I already knew they were equivalent, and I in turn pointed out that switching from "if P then Q" to "if Q then P" makes all the difference; I never objected to "if Q then P", only "if P then Q", because the former is equivalent to falsification, and the latter is the only kind of confirmationism I'm against here).

So if you're affirming that the good direction of inference is from "if Q then (probably) P", and "(probably) Q", to "(probably) P", and you're not saying that the inference from "if P then (probably) Q", and "(probably) Q", to "(probably) P", is a good inference, then you're completely agreeing with me, even if you think you're not.

the correct formulation for confirmationism is "if Q then P", — Janus

I just went to look up a quote about confirmationism in a reliable source, the Stanford Encyclopedia of Philosophy, to settle this merely nominal dispute once and for all, and I found something interesting: there are two mutually contradictory things both called "confirmationism".

I learned confirmationism in school as synonymous with the hypothetico-deductive method, which to quote SEP means:

e HD-confirms h relative to k if and only if h ∧ k ⊨ e and k ⊭ e; — SEP

Where e is some evidence, h is some hypothesis, k is some set of background assumptions, and ⊨ is a symbol for entailment, i.e. necessary implication.

So on that hypothetico-deductivist confirmationist account, if the hypothesis (and background assumptions) entail the evidence (but the background assumptions alone don't), then seeing that evidence confirms the hypothesis. That is exactly what I have been saying confirmation claims, except using "P" for the hypothesis+assumptions together, and "Q" for the evidence: that if P implies Q, and Q is the case, then P is the case.

But, it seems, Hempel's model, which I learned in school as something against confirmationism and more a step toward falsificationism -- because it says exactly the opposite of that hypothetico-deductivism above, just like falsification does -- is apparently also called "confirmationism", and I'm guessing that that's where you're coming from, Janus. Hempel's account says:

e confirms h relative to k if and only if e disconfirms ¬h relative to k. — SEP

Which is pretty much the same thing falsification says, except it doesn't call that "confirming" h, because falsificationism uses the term "confirmation" to refer to hypothetico-deductive confirmation. It just says that that's falsifying ¬h, which of course entails that h as well.

I get the feeling that maybe your point is that zero probability is not quite identical to impossibility, just very close. But even given that that's the case, it's not counter to my main thrust here, since as explained to Janus above, I'm not depending on these implications I'm loosely using here being completely strict absolute deductions; the same points I'm making apply even if they're taken only to be probabilistic relationships, as I'll explain more to Srap below right now.



Bayes theorem is:
P(H | E) = P(E | H) * P(H) / P(E)

Some simple algebra can rearrange that to:
P(H | E) * P(E) / P(E | H) = P(H)

Things like “P(X | Y)” are often phrased as “the probability of X given Y”, but that means the same thing as “the probability that X is true if Y is true” or “the probability that if Y is true then X is true” or “the probability that Y implies X”. “X given Y” = “X if Y” = “if Y then X” = “X implies Y”. It’s all the same thing. Just wrap a “the probability that” around any of those and you have what “P(X | Y)” means.

So we can rephrase the above formula as:

P(E implies H) * P(E) / P(H implies E) = P(H)

Meanwhile, the standard form of a falsificationist inference is:

~H implies ~E, and E, therefore H.

or equivalently:

E implies H, and E, therefore H.

If we want to make that probabilistic instead, using formal notation instead of just sticking “probably”s in there as I’ve been doing because I assumed we were all smart people who can read between the lines, we’d then have:

P(E implies H) * P(E) = P(H)

or if you like:

P(H | E) * P(E) = P(H)

The only difference between that and Bayes formula is that Bayes also has P(E | H) in the denominator on the left (of the rearranged presentation), i.e. the probability of the hypothesis is also inversely proportional to the probability that the hypothesis implies the evidence, i.e. the probability of the evidence given the hypothesis.

In other words, not only does Bayes theorem say that the more probably that the evidence implies the hypothesis, the more probable the hypothesis is true — which is exactly what falsification says, since “E implies H” is equivalent to “~H implies ~E” — additionally, the more probably that the hypothesis predicts the evidence, the less probable that the hypothesis is true — the exact opposite of what confirmationism would have you think.

That is to say, the standard form of a confirmationist inference:
H implies E, and E, therefore (probably) H.

rendered into proper probabilistic notation would be:
P(H implies E) * P(E) = P(H)

or if you like:

P(E | H) * P(E) = P(H)

where the first term there is exactly what Bayes theorem has the inverse of, in addition to also having the converse (“P(E implies H)” or "P(H | E)").

Let me repeat all that. Falsification rendered probabilistically:

P(H | E) * P(E) = P(H)

Bayes theorem, algebraically rearranged:

P(H | E) * P(E) / P(E | H) = P(H)

Versus confirmationism rendered probabilistically:

P(E | H) * P(E) = P(H)

So if anything, Bayes theorem is doubling down on falsificationism vs confirmationism, besides just rendering the whole thing into a probabilistic form.

Things like “P(X | Y)” are often phrased as “the probability of X given Y”, but that means the same thing as “the probability that X is true if Y is true” or “the probability that if Y is true then X is true” or “the probability that Y implies X”. “X given Y” = “X if Y” = “if Y then X” = “X implies Y”. It’s all the same thing. Just wrap a “the probability that” around any of those and you have what “P(X | Y)” means. — Pfhorrest

We shall see that there is no way to interpret a conditional connective so that, with sufficient generality, the probabilities of conditionals will equal the appropriate conditional probabilities. — David Lewis

The bar in ‘pr(H | D)’ is not a connective that turns pairs H, D of propositions into new, conditional propositions, H if D. — Richard Jeffrey

Maybe you're right, @Pfhorrest; but maybe Lewis and Jeffrey are right. Don't care.

I won't have time to read both of those in their entirety tonight, but just reading the opening of the Lewis paper, he also says:

I shall take it as established that the assertability of an ordinary indicative conditional A -> C does indeed go by the conditional subjective probability P(C|A). — David Lewis

Which seems to be the same thing as I'm saying. I don't yet know how he reconciles that with the bit you've quoted. I suspect that perhaps this will turn out to be a roundabout way of saying that natural language conditionals don't mean exactly what material implications mean, i.e. a natural language assertion of "if P then Q" doesn't just mean "not(P and not Q)", but instead it means something more like the "(Q|P)" in probabilistic notation. If so then that's fine with me, as I'm using conditionals in a natural language way here, not especially tied to them meaning precisely what material implication means.

ETA: Yep that looks like it, as just below your quote from Jeffrey he writes of Lewis' work:

That is David Lewis’s “trivialization result”. In proving it, the only assumption needed about ‘if’ was the eqivalence (2) of ‘If X, then if C then D’ with 'If X and C, then D'. — Richard Jeffrey

Which equivalence hinges on the conditionals being taken as material implications.

Things like “P(X | Y)” are often phrased as “the probability of X given Y”, but that means the same thing as “the probability that X is true if Y is true” or “the probability that if Y is true then X is true” or “the probability that Y implies X”. “X given Y” = “X if Y” = “if Y then X” = “X implies Y”. It’s all the same thing. Just wrap a “the probability that” around any of those and you have what “P(X | Y)” means. — Pfhorrest

If you're willing to "pass up" the conceptual hierarchy to more everyday language use of the concepts, I think that's fine. So long as you're aware you're talking about different (but related in some way) concepts than material implication and probability.

I was busy for a few days and this thread seems be going nowhere so I'm sorry if this reply is no longer relevant, but I didn't want to just leave it hanging.

1. Beliefs are not propositions. Beliefs are states of mind equivalent to a tendency to act as if... — Isaac

Would this mean then that animals have beliefs? — Coben

I'm not a purist about language except within context. 'Beliefs' can obviously mean a range of things depending on how we're using the word, so I want to be clear that when I say beliefs are 'tendencies to act as if...' I mean that in the context of psychology. As such, intelligent animals can have beliefs, but machines can't, simply by definition (beliefs are something which minds have). In a functional sense, I'm quite happy to see a belief as no different to the tendency of a thermostat to turn the radiator down when the room is to hot, but that wouldn't be a belief - despite it's functional similarity - because it's not a state of a mind.

a) not possible to have a belief which is contrary to the evidence of your senses (beliefs are formed by a neurological process of response to stimuli), and — Isaac

Does this mean that one cannot come to believe things that are counterintuitive: relativity, for example, or that the earth actually revolves around the sun. If we take the latter case that we can find empirical evidence that this is the case, very few people actually do that. Or that color exist outside us. — Coben

No, because there's no obligation on our senses to deliver us a coherent set of data. We can observe the sun rise and set, we can observe pictures of the earth from space, we can listen to scientists whom we trust talk about orbits and construct a mental image of such - all these may well range from slightly inconsistent to completely incompatible. We can believe one at some time and another at another, we can believe one in one context and another in another. Nothing enforces coherence.

-- this leads to the more general criticism that there is no target of the normative claim, it's like telling people that they ought to breathe. — Isaac

What was his normative claim? — Coben

The actual normative claim seems to be that we should reject a system of beliefs which, in it's entirety, is inconsistent, in favour of one which is consistent, but that we should not prefer one consistent set of beliefs over another. so long as they are consistent, they're OK.

Since I doubt anyone would agree that we should maintain an inconsistent belief system, or that we can dismiss the beliefs of others on grounds other than consistency, I don't see how there's any proper target here.

The problem is, @Pfhorrest only resorts to this wider sense of the claim when pushed, as soon as that pressure is released we go back back to the much more narrow sense - that some people hold beliefs which (my personal take on) empirical evidence shows to be wrong, and they shouldn't.

Like, I'm going to out on a limb and say the vast majority of these sorts of philosophies, they're looking for a stick with which to beat their moral or ideological opponents, and no stick is bigger or heavier than "the world says you're wrong", or "logic says you're wrong". Unfortunately the world turns out to be fiendishly complicated and if it says anything at at it's in virtually indecipherable code so these projects always fail.