Popper - Falsification and a Life's Struggle

[magellan004]

This thread is to examine the concept of Falsification as presented by Karl Popper in the essay Science as Falsification http://nsmserver2.fullerton.edu/depa...web/Popper.htm .The speech excerpt is fairly short and easy to read. In that speech Popper refers to a theory of Einstein to do with bending of light and a related ‘test’ carried out by Arthur Stanley Eddington. Details of the Eddington expedition are outlined in Einstein, Eddington and the 1919 Eclipse by Peter Coles http://arxiv.org/pdf/astro-ph/0102462v1.pdf .

Karl popper was born in 1902. The events he describes in the above essay took place around 1919 and he delivered his speech around 1963. So after the passage of 44 years and a life spent contemplating philosophical ideas Science as Falsification might be seen as more of an exploration into how Popper was profoundly affected as a teenager and his quest to come to terms with his own struggle with his affectations – which remained unaltered over the passage of time.

But for this thread I will concentrate on the logic of his concept of Falsification.

All contributions are welcome and there is only one rule and it is mainly for Tiggy – Be well mannered, use people’s ID’s, do not insult or ridicule.

First – the groundwork:
True and False
Truth is invariable and non-contradictory.
IF ‘2+2 = 4’ is true today then it was true yesterday and will be true tomorrow.
IF ‘2+2 = 4’ is true then ‘2 + 2 = 5’ is False.
The Statement ‘A’ ‘2+2=5 is False’ can only be True under one condition – that A is a true statement.
IF the Statement ‘This swan is White’ is refuted by ‘I observe that this swan is black’ then the statement ‘I observe this swan is black ‘ is true.
The Statement ‘This swan is White’ is not refuted by an observation by John Smith – a chronic liar and fraud, lying and stating ‘I observe … ‘.
To refute a claim requires a true statement.

As used by Popper –‘refute’, falsify’ and ‘show false’ are equivalent terms.

To refute a statement is to show the statement is false.
Only a true statement can refute a claim.
Observation as Refutation
Getting back to Popper, he says ‘If observation shows that the predicted effect is definitely absent, then the theory is simply refuted.’.
We know that only a true statement can refute so if a report of Observation Y (say from Eddington) refutes claim A then Observation Y must be true. If Observation Y was provisional, conditional or qualified then it has no power to refute.

What do you think?

Coming up – The Empirical Process and Testing

Magellan

[ZackMartin]

I think the logical positivists put the emphasis on verification. Popper was an interesting switch from that, putting the emphasis on falsification instead. I think both have value, and both were wrong to privilege one to the exclusion of the other.

[phank]

I think the logical positivists put the emphasis on verification. Popper was an interesting switch from that, putting the emphasis on falsification instead. I think both have value, and both were wrong to privilege one to the exclusion of the other.

They are very different beasts. Falsifiability is the same thing as testability. Anything that can be tested, is in principle capable of failing the tests and being falsified, and anything that is falsifiable can ONLY be falsified through testing. So testability is a necessary characteristic of any theory. And it's important to bear in mind that we're speaking here strictly of the world of physical observation, NOT about artifical formal systems like math or much of religion.

Verification is something different. It is possible for something to be testable, but unverified (because the test has not been performed), and it is possible for something to be correct but not testable. Whether something is correct or not is different from whether something has been or can be ESTABLISHED to be correct or not.

(As a historical note, Popper first put forth his ideas of falsification, and they were found to be untenable. He eventually modified his position to be more realistic. So it's important to be careful when someone writes that "Popper said" something. Like any reasonable person, he changed his mind when his ideas didn't work.)

[ZackMartin]

Let's take the claim "When you die, you will enter into a conscious afterlife of some shape, form or description." Is there an experiment which can verify that claim? Sure there is; here's a gun, shoot yourself in the head. If you find yourself in heaven, hell, purgatory, limbo, wandering around as a ghost, being reincarnated, or in any other such destination, then the claim has been verified by experiment. If you death is followed by nothingness, your won't have any such verificational experiences, and thus the claim will not be verified. Even though it will be false, you won't be aware of its falsehood, so it cannot be said to have been falsified. So, interestingly, this claim is a claim that can be experimentally verified, but cannot be experimentally falsified.

Now let's take the claim "When you die, you cease to exist". We can perform the same experiment. If you find yourself in an afterlife of some form, the claim is falsified; but if you cease to exist, the claim is not verified (since you don't exist to verify it). So here by contrast is a claim that can be experimentally falsified, but cannot be experimentally verified.

(It does not need to be said that I do not advocate anyone actually carrying out such an experiment of shooting themselves in the head in order to find out whether an afterlife exists; it theoretically counts as an experiment, but it is not an experiment I advocate anyone carry out in practice.)

[magellan004]

They are very different beasts. Falsifiability is the same thing as testability. Anything that can be tested, is in principle capable of failing the tests and being falsified, and anything that is falsifiable can ONLY be falsified through testing. So testability is a necessary characteristic of any theory. And it's important to bear in mind that we're speaking here strictly of the world of physical observation, NOT about artifical formal systems like math or much of religion.

Verification is something different. It is possible for something to be testable, but unverified (because the test has not been performed), and it is possible for something to be correct but not testable. Whether something is correct or not is different from whether something has been or can be ESTABLISHED to be correct or not.

(As a historical note, Popper first put forth his ideas of falsification, and they were found to be untenable. He eventually modified his position to be more realistic. So it's important to be careful when someone writes that "Popper said" something. Like any reasonable person, he changed his mind when his ideas didn't work.)

'Falsifiability is the same thing as testability.' In his essay Science as Falsification Popper does say that.
This would be correct if a test always leads to a conclusion of True/False. Given that the history of science is also a history of many mistakes, this connection does not seem to hold. Popper's Falsifiability is about True/False.

However Popper also says that refutation occurs by the act of observation (as opposed to a test being conducted). So his message seems mixed.

About Popper changing his mind - That may be true. When I write 'Popper said' I am referring to the article by Popper which I referred to as the foundation of this thread in the opening post. The article Science as Falsification interested me because after 40 odd years Popper had not changed his mind on Falsification. Indeed he is explaining why his early idea was correct. It may be that he later changed his mind. However by that time many people had adopted his idea of falsification as gospel.

Magellan

[magellan004]

Let's take the claim "When you die, you will enter into a conscious afterlife of some shape, form or description." Is there an experiment which can verify that claim? Sure there is; here's a gun, shoot yourself in the head. If you find yourself in heaven, hell, purgatory, limbo, wandering around as a ghost, being reincarnated, or in any other such destination, then the claim has been verified by experiment. If you death is followed by nothingness, your won't have any such verificational experiences, and thus the claim will not be verified. Even though it will be false, you won't be aware of its falsehood, so it cannot be said to have been falsified. So, interestingly, this claim is a claim that can be experimentally verified, but cannot be experimentally falsified.

Now let's take the claim "When you die, you cease to exist". We can perform the same experiment. If you find yourself in an afterlife of some form, the claim is falsified; but if you cease to exist, the claim is not verified (since you don't exist to verify it). So here by contrast is a claim that can be experimentally falsified, but cannot be experimentally verified.

(It does not need to be said that I do not advocate anyone actually carrying out such an experiment of shooting themselves in the head in order to find out whether an afterlife exists; it theoretically counts as an experiment, but it is not an experiment I advocate anyone carry out in practice.)

I agree that there seems to be no good reason to say that as a methodology tests or observations can show one thing (eg truth) but not the other (eg false). The history of science mistakes tells us that neither 'Show true' nor 'Show false' happens as a result of either methodology.

When you use the term 'Verify' do you mean 'To show true'? Popper defines Falsify as Show False but he does not define verify.

Magellan

[ZackMartin]

When you use the term 'Verify' do you mean 'To show true'? Popper defines Falsify as Show False but he does not define verify.

Yes.

Verify = obtain sufficient evidence to rationally conclude that a proposition is true.

Falsify = obtain sufficient evidence to rationally conclude that a proposition is false.

The standard of evidence required is not absolute certainty - absolute certainty is something which science neither provides nor requires.

[phank]

'Falsifiability is the same thing as testability.' In his essay Science as Falsification Popper does say that.
This would be correct if a test always leads to a conclusion of True/False. Given that the history of science is also a history of many mistakes, this connection does not seem to hold. Popper's Falsifiability is about True/False.

As usual, this is a misunderstanding. In the empirical world, models generate testable hypotheses. But NO test can show that the hypothesis is true. It can only add support to the hypothesis. A test can show the hypothesis is false, but never true. So the test leads to a conclusion of False/not proved false yet. Popper was well aware that it is not possible to PROVE anything in science. The best that science can do, in the words of Stephen Jay Gould, is to support a proposition to a degree where it would be perverse to withhold provisional assent.

However Popper also says that refutation occurs by the act of observation (as opposed to a test being conducted). So his message seems mixed.

But now you understand that it seems mixed only because you misunderstood it. Observation can show a hypothesis to be false, and refute it. But at best it can only lend support to a hypothesis. In science, it is inherently the case that the NEXT observation, of anything, might refute any theory or hypothesis. And this condition holds indefinitely.

About Popper changing his mind - That may be true. When I write 'Popper said' I am referring to the article by Popper which I referred to as the foundation of this thread in the opening post. The article Science as Falsification interested me because after 40 odd years Popper had not changed his mind on Falsification. Indeed he is explaining why his early idea was correct. It may be that he later changed his mind. However by that time many people had adopted his idea of falsification as gospel.

Magellan

Popper honed his ideas over time, but the core is that if a claim cannot be tested in principle, it is not a scientific claim. Yes, granted science has plenty of claims that CAN be tested in principle but would require techology far beyond current capabilities. You know, colliders the diameter of the galaxy, things like that. But perhaps it would help you to use the term testability rather than falsification.

[phank]

Yes.

Verify = obtain sufficient evidence to rationally conclude that a proposition is true.

Only for trivial propositions, not for theories. If your proposition is that if you drop this brick it will fall to the ground, you can verify that it does. But if your proposition is that all dropped objects will always fall to the ground, this cannot be verified. The very next object you drop might not fall.

Falsify = obtain sufficient evidence to rationally conclude that a proposition is false.

This is always possible.

The standard of evidence required is not absolute certainty - absolute certainty is something which science neither provides nor requires.

In the empirical world, absolute certainty is a delusion. As in law, the standard of evidence is reasonable doubt.

[ZackMartin]

Only for trivial propositions, not for theories. If your proposition is that if you drop this brick it will fall to the ground, you can verify that it does. But if your proposition is that all dropped objects will always fall to the ground, this cannot be verified. The very next object you drop might not fall.

What made us believe that general relativity was true? Because numerous of its predictions were experimentally verified.

Verification drives science as much as falsification does. Falsifying competing theories by itself is not a sufficient reason to accept a theory; one also needs verification of it. Not perfect verification, not absolutely certain verification, but at least some verification is required. If a scientific theory is completely unverified, we should not accept it, even if all of its competitors have been falsified. Even if we falsify a theory's competitors, we can't presume that the theories we know about are the only possible theories; maybe the so far unfalsified theory will be falsified too, and then someone will devise a new theory that no one has even thought of yet.

[phank]

What made us believe that general relativity was true? Because numerous of its predictions were experimentally verified.

Not exactly. Each verified prediction supports the threory, but none can "prove" that it's true. And it is known to be incompatible with quantum mechanics.

Verification drives science as much as falsification does. Falsifying competing theories by itself is not a sufficient reason to accept a theory; one also needs verification of it. Not perfect verification, not absolutely certain verification, but at least some verification is required.

These are basically the same coin. A hypothesis makes a testable prediction. If the prediction fails, the hypothesis is falsified. If the prediction is correct, the hypothesis is supported. Support and verification are somewhat different things, and this depends on the scope and nature of the hypothesis as I tried to illustrate.

If a scientific theory is completely unverified, we should not accept it, even if all of its competitors have been falsified.

Um. A scientific theory isn't a single hypothesis, it's a fairly broad-ranging explanation of a fairly extensive set of observations, and relies on a good many verified or supported hypotheses. A theory is NOT a hunch, guess, or speculation. Ultimately, a theory is a "best-fit" explanation for a large set of related observations.

Even if we falsify a theory's competitors, we can't presume that the theories we know about are the only possible theories; maybe the so far unfalsified theory will be falsified too, and then someone will devise a new theory that no one has even thought of yet.

In principle, sure. But it's very rare that a scientific theory is falsified entirely. Instead, theories are extended and refined. Hypotheses are falsified all the time. You're right that where the number of potential competing explanations is large, eliminating one or even all but one so-far-presented explanation can never establish an explanation by default. Explanations MUST have empirical support. And you're also correct that even those must always be regarded as provisional. It's one of the tenets of science that all theories are forever subject to modification as new observations are made.

[ZackMartin]

Not exactly. Each verified prediction supports the threory, but none can "prove" that it's true.

Science doesn't deal with the certainties of mathematical proofs. "Verification" in science doesn't mean "proven absolutely true", just "shown to be significantly more likely than not to be true".

Even falsification is probabilistic. If you got experimental data that appears to falsify a theory, there is still a possibility that your data is due to experimental error, and the theory is in fact true. With repeated experiments, you can reduce that probability down as far as you want, but it never reaches zero.

And it is known to be incompatible with quantum mechanics.

Sure, but that doesn't mean it isn't verified. To be accepted, a scientific theory doesn't have to be perfect, just generally useful. The fact is that it can make useful predictions most of the time. Even if we one day replace general relativity with some new "quantum gravity" theory or "string theory" or whatever, we'd have no reason to suppose that theory was absolutely perfect either, just closer to perfection than Einstein's is.

These are basically the same coin. A hypothesis makes a testable prediction. If the prediction fails, the hypothesis is falsified. If the prediction is correct, the hypothesis is supported. Support and verification are somewhat different things, and this depends on the scope and nature of the hypothesis as I tried to illustrate.

If the prediction fails, the falsehood of the hypothesis is supported but not proven. You could only say it had been definitely disproven, if you had reduced all sources of error to zero, and you can't do that.

[phank]

Science doesn't deal with the certainties of mathematical proofs. "Verification" in science doesn't mean "proven absolutely true", just "shown to be significantly more likely than not to be true".

Even falsification is probabilistic. If you got experimental data that appears to falsify a theory, there is still a possibility that your data is due to experimental error, and the theory is in fact true. With repeated experiments, you can reduce that probability down as far as you want, but it never reaches zero.

Yep. I think this needs to be said straight out, as you have done. Tentative scientific conclusions rest on the weight of evidence, and it's unusual for any single finding to be dispositive of anything by itself.

Sure, but that doesn't mean it isn't verified. To be accepted, a scientific theory doesn't have to be perfect, just generally useful. The fact is that it can make useful predictions most of the time. Even if we one day replace general relativity with some new "quantum gravity" theory or "string theory" or whatever, we'd have no reason to suppose that theory was absolutely perfect either, just closer to perfection than Einstein's is.

Yes, I agree. I'm concerned with the connotations of the word. The dictionary defines "verified" as "proven to be true", and the thesaurus gives the synonym of "verified" as "proven". So this is why I try to avoid that word - it's misleading in this context. I prefer "strongly supported". When someone replicates a study and gets the same results, generally the scientific literature doesn't say the second study "verified" the first, but rather that the second study "replicated the results of the first."

If the prediction fails, the falsehood of the hypothesis is supported but not proven. You could only say it had been definitely disproven, if you had reduced all sources of error to zero, and you can't do that.

In principle, this is correct. In practice, this is generally not an issue. Well constructed experiments provide very clear operational definitions - that is, unambiguous predictions implied by the original model. Again, consider the above example, where the hypothesis was that all dropped objects will fall. It takes only ONE object that fails to fall, to falsify this hypothesis. There is no solipsistic wrangling over whether the hot air baloon "might have fallen, and all observations that it did not MIGHT have been mass hypnosis" or some such. At some point, observation must be given provisional assent or the concept of observation becomes meaningless.

Which isn't to say that despite clear definitions, results might not be ambiguous. They often are. But when they are not, they are not. Once you start wandering into the no-mans-land of Absolutely vs. Clearly Unambiguous, you're really into philosophy and not science.

[ZackMartin]

In principle, this is correct. In practice, this is generally not an issue. Well constructed experiments provide very clear operational definitions - that is, unambiguous predictions implied by the original model. Again, consider the above example, where the hypothesis was that all dropped objects will fall. It takes only ONE object that fails to fall, to falsify this hypothesis. There is no solipsistic wrangling over whether the hot air baloon "might have fallen, and all observations that it did not MIGHT have been mass hypnosis" or some such. At some point, observation must be given provisional assent or the concept of observation becomes meaningless.

Which isn't to say that despite clear definitions, results might not be ambiguous. They often are. But when they are not, they are not. Once you start wandering into the no-mans-land of Absolutely vs. Clearly Unambiguous, you're really into philosophy and not science.

But very few real hypothesises will be as simplistic as "dropped objects will fall". If you drop an object and it doesn't fall, is gravity disproven? Or maybe the object is light, and there is a strong upward wind current, etc?

Many hypothesises are tested with statistical methods. These methods are subject to sampling error. As the sample size increases, the sampling error falls, but it never reaches zero. So regardless of whether your conclusion is positive or negative, you can never say it is absolutely certain, but with more work you can make it as close to certain as you like, but in science, certainty is a destination one can never reach no matter how close one approaches to it, and that is just as true for evidence against a theory as it is for evidence for it.

[phank]

Well, we're not disagreeing here. My argument would be that eventually you reach a point of diminishing returns, where enough slightly different experiments have been done to pretty solidly establish that something does or does not happen (within experimental constraints), and you move on. And you don't revisit it unless and until you start getting results incompatible with what you thought was established, and you must return and revisit it through new eyes. So we seem to agree that this is an iterative process producing provisional results.