Should the concept of Falsifiability be abandoned in science?

[wolram]

http://arxiv.org/pdf/1504.00108.pdf

Some of the most obviously correct physical theories – namely string theory and the multiverse – make no testable predictions, leading many to question whether we should accept something as scientific even if it makes no testable predictions and hence is not refutable. However, some farthinking physicists have proposed instead that we should give up on the notion of Falsifiability itself. We endorse this suggestion but think it does not go nearly far enough. We believe that we should also dispense with other outdated ideas, such as Fidelity, Frugality, Factuality and other “F” words. And we quote a lot of famous people to support this view.

Surly this paper is nonsense.

 

[Borg]

A little late for April First.

 

[e.bar.goum]

That was pretty excellent. Thanks for the laugh! My favourite part was the reference list.

[18] Adams D, 1979, ‘The Hitchhiker’s Guide to the Galaxy’, Pan
[19] Not really relevant, but we realized that we hadn’t quoted Einstein yet.
[20] Actually, that’s a better quote, so ignore the last one.

 

[Chronos]

It appears appropriate for April 1.

 

[MarcusAgrippa]

Don't jeer at this paper or write it off as a joke! Well - actually - this one is a joke - but here is what is not a joke: there are creationists out there who are actually advocating that many truths are unfalsifiable and that we should reject falsifiability from our world view and accept as truth things that are not falsifiable. A convenient untruth?

 

[wabbit]

You may be right but I am inclined to not look beyond the more explicit target here. Creationists might advocate that some evident truths are above falsifiability, but they do not seem to have a lot of influence within Science.

 

[wolram]

More seriously how much if any thing is falsifiable in string theory, and what about LQG?

 

[wabbit]

Actually, I'm not sure "falsifiability" is such a great criterion, because I don't believe scientific theories are either true nor false - they can be consistent, testable, predictive, useful, powerful, etc... As to ST and LQG, it seems early, can either even properly be called a theory, rather than a tentative theory, at this stage ?

At some point some proponents ST appeared to be tempted to abandon criteria of predictivity, testability, and usefulness, so what would be developped under such a program is less likely to be very good at any of those. But this may have been straightened out now.

 

[diogenesNY]

The name of the primary author is an anagram of 'April Fool'.

diogenesNY

 

[Garth]

The name of the primary author is an anagram of 'April Fool'.

diogenesNY

But don't write him off, Ali Frolop is a serious author with at least five eprints to his name, together with Douglas Scott they have published:
Cosmic Conspiracies, 2 April 2006,
Natural Dark Energy[Natural Dark Energy ,30 March 2007,
Down-sizing Forever, 31 March 2008,
Ali Frolop then went into retirement but came out to write The CMB flexes its BICEPs while walking the Planck, 31 March 2014, because he was so excited by the "swirly patterns" in the CMB,
and then of course A Farewell to Falsifiability, 1 April 2015, which included the insightful comment not noted anywhere else:

[10] Popper K., 1959, ‘The Logic of Scientific Discovery’ Routledge.
[11] However, Popper’s skepticism towards astrology is just what you’d expect from a Leo.

2 April 2015
Garth

 

[bapowell]

Actually, I'm not sure "falsifiability" is such a great criterion, because I don't believe scientific theories are either true nor false - they can be consistent, testable, predictive, useful, powerful, etc...

Really? So if you have a theory that predicts that all swans are white, and then I observe a black swan, your theory isn't falsified?

 

[wabbit]

Sure, you're right a theory can be true or false in principle, and in some cases this may be a criterion, but in practice is that an effective criterion?

In the case of the theory of swans, it would remain valid with an exception, so it would be slightly amended with a footnote. Actually, "all swans are white" is a pretty good approximation, and that's the theory we can use in most cases, falsified or not.

But perhaps I am overstating the case, I am not sure. Still, predictive power etc. seem to provide a more effective set of selection criteria. Would you have examples where falsifiability is a better choice?

Also, I suspect most theories do not make absolute logical statements, they describe how to obtain useful approximations within a certain domain - so falsifiability isn't really applicable, or at least it's not effective. If you say "this theory is falsified, it claimed 1% precision and I showed here a 1.1% deviation", how powerful is that falsification?

 

[bapowell]

But perhaps I am overstating the case, I am not sure. Still, predictive power etc. seem to provide a more effective set of selection criteria. Would you have examples where falsifiability is a better choice?

Predictive power is certainly important, but I don't see why you think it is set apart from, or mutually exclusive with, falsifiability. There are essentially two ways of resolving a hypothesis: verification or falsification. Either requires that the hypothesis be predictive. Now, the more predictive it is, the more falsifiable/verifiable it is. Karl Popper (the guy who popularized falsification as *the* criterion separating science from the riff raff) spoke of corroborating a theory -- any theory that survived rigorous testing could be said to be corroborated. The degree of corroboration depends on the severity of the test as well as the predictiveness (akin to the empirical content) of the theory/hypothesis. So he viewed falsification as a means of "killing off" weak or bad hypotheses, in effect, he viewed the progress of science as a selection principle among conjectures that worked to single out a well-corroborated theory that survived the trials, and predictiveness was one characteristic of fitness.

Also, I suspect most theories do not make absolute logical statements, they describe how to obtain useful approximations within a certain domain - so falsifiability isn't really applicable, or at least it's not effective. If you say "this theory is falsified, it claimed 1% precision and I showed here a 1.1% deviation", how powerful is that falsification?

Sure, we can usually only falsify by degrees. But there are several cases where falsification was essentially an open and shut case. Consider Fred Hoyle's steady state model of the universe. This model could be said to predict no CMB and no Big Bang Nucleosynthesis. That these things exist quite directly contradict the steady state model -- the only uncertainty in the falsification would arise from the uncertainty of the detections, which in the case are vanishingly small.

 

[Almeisan]

Physics abandoning falsifiability is no joke.

 

[wabbit]

I don't disagree with that, though I'm still not sure why falsifiability, a concept derived from logical truth (though modified from that, so it is a bit of a misnomer - progressive corroboration is a rather different concept), is such a great concept, or what it really adds. It is often publicized as a great breakthrough and a key test, I am skeptical of that.

And in the case of steady state, we do have something pretty close to a categorical statement, so I won't pretend it wasn't falsified : )

 

[wabbit]

Physics abandoning falsifiability is no joke.

It is no joke I agree, because under that heading what we re talking about goes much further than that - abandonning testability, predictive power, etc... So falsifiability here is a convenient shorthand for that.

And I am probably just arguing semantics here, focusing on a word which by itself is not the issue - in practice when we say "abandoning falsifiability" we don't mean abandonning a specific criterion, but abandonning the scientific method (which existed well before that word of course, and proceeded pretty much the same after Popper as before I suspect).

Edit : in the end perhaps "falsifiabilty" is a useful word in fact, precisely because it provides an easily communicated one-word proxy for "the scientific method", and it's convenient to use.

 

[diogenesNY]

Interesting how an April Fools event has stimulated a very interesting discussion in re Philosophy of Science.

I think it is worth noting that Big Bang and Steady State were not simple concise hypotheses that could be invalidated by a simple negative proof, but rather were and are during their respective useful lifetimes big messy, multi tentacled models that were subject to ongoing development, tweaking revision and addition as more and more observation and data about the observable universe continued to to be acquired. There were many missteps and inaccurate predictions along the way as things developed, but incompleteness does not imply falsification. Supernova based nucleosynthesis was originally championed by Hoyle as supporting steady state. I t turned out to be an excellent bit of astrophysics that fit quite nicely with the observed universe and ended up meshing quite nicely with Big Bang generally and the formation of structures within the observable universe more specifically.

I guess what I am getting at is that straight forward falsifiability is just fine when one is working on fairly discrete and specific hypothesis. With the bigger, shaggier and more slowly developed models, things evolve over time as more observations are made, incompleteness is reduced and overall perspectives must necessarily be revised in tune with the ongoing research programs and as models are further developed and made ever more complete.. While I am a big fan of Popper, perhaps for some of the bigger and messier issues, particularly as applied to broad research programs that take place over a good deal of time, one might well also look to someone like Thomas Kuhn (e.g. _Structures of Scientific Revolutions_ Columbia University Press 1968) for some insight into how we get (or got) from A to B.

diogenesNY

 

[bapowell]

I don't disagree with that, though I'm still not sure why falsifiability, a concept derived from logical truth (though modified from that, so it is a bit of a misnomer - progressive corroboration is a rather different concept), is such a great concept, or what it really adds. It is often publicized as a great breakthrough and a key test, I am skeptical of that at the very least. (I am not saying that unfalsifiable statements should be accepted).

Because falsification is the only logically sound means of hypothesis testing (it uses modus tollens): if $H\rightarrow O$ then $\tilde{O} \rightarrow \tilde{H}$, where the tilde should be read as 'not'. Meanwhile, the act of verification is known to commit the fallacy of affirming the consequent: if $H\rightarrow O$ then $O \rightarrow H$. Further, the verification procedure relies on induction which we know suffers from well-known problems. Popper was simply looking for a logically valid, non-inductive mode of inference and demarcation.

 

[wabbit]

OK the ability to produce falsifiable statements (which seems to be the same as testability) is important, even though being falsified in this sense doesn't necessarily kill a theory, it can simply reduce its domain of validity (as in Newtonian Gravity's prediction for Mercury's perihelion).

 

[bapowell]

OK the ability to produce falsifiable statements (which seems to be the same as testability) is important, even though being falsified in this sense doesn't necessarily kill a theory, it can simply reduce its domain of validity (as in Newtonian Gravity's prediction for Mercury's perihelion).

Indeed. And I think we all agree that in reality, falsification is maybe not so cut and dry. It's a messy business.

 

[Nugatory]

A bit of discussion of this paper towards the end of this thread: https://www.physicsforums.com/threads/quantum-music.806169/

 

[Garth]

Physics abandoning falsifiability is no joke.

I agree, we had a thorough discussion on the subject in the thread You think there's a multiverse? Get real .

Both String Theory and the Multiverse conjecture are not only difficult or impossible to robustly test but also the throw up such a large number of models (10⁵⁰⁰ is quoted for String Theory) that one model may be found to fit any observation.

The problem with theories that can explain everything and anything is that they actually explain nothing.

A serious subject from an April Fool's joke!

Garth

 

[Chronos]

My discomfort with multiverse theory is it attempts to dismiss fine tuning 'problems' based on anthropic pink fairies, or obscenely improbable statistics. That looks more like mythology than science to me.