dgruss23 introduced falsification in this Q&A thread, post #32:I'm curious about what BAUT members think of the role of falsification in modern science, particularly those parts of science directly relevant to the scope of BAUT (astronomy, cosmology, planetary sciences, astrobiology, ....).

My own view is summarised in my response to drguss23's post:In the very next post, dgruss23 continues:Now I do NOT want to discuss the case for or against MOND, or collisionless CDM, or ... (although they may be used as examples); I am curious to know what others think of the relevance of falsification in modern science.

For example, does it describe how scientists 'do' science, even if only more-or-less? Or is it an unscientific notion, if only because it itself cannot be falsified (even in principle)?

In addition to Mercury's anomalous perihelion*, let's consider the neutrino: the initial (beta decay?) observations could have been said to 'falsify energy conservation' (and, AFAIK, some folk said exactly that, despite being quite familiar with Noether's theorem). Several decades later a testable hypothesis ('neutrino') that was built on energy conservation was confirmed, by Reines and Cowan. For those involved in researching Mercury's orbit, or beta decay (and other neutrino-related stuff), prior to GR/{Reines and Cowan}, how helpful would statements such as "Newton's universal law of gravitation has been falsified" and "the law of conservation of energy has been falsified" have been?

And in the case of collisionless CDM and MOND, how helpful is it that MOND was falsified before it was even written down^?

* which remained 'anomalous' for how many decades? IIRC, at least the working life of an average scientist, twice over!

^ MOND is inconsistent with special relativity (SR), and SR has passed an enormous range of experimental and observational tests.

To say that "Newton's law of gravitation has been falsified." Not very useful at all. Most scientist know that they dont know everything. At least that was the way I was taught and the conclusion all my classmates came up with.

To say something is 'falsified' does not mean it is useless and should be tossed. Newtonian dynamics is virtually dead on for everything we can measure that doesnt involve a close approach to a stellar mass. As a side note, there should have been anomalous changes in sun grazing comets also. When you combine all the data, it shows that there is some effect due to mass on how orbits are shaped. Oddly enough, then someone put that bit of data into a theory, added in SR, and POOF, we had a pretty good explanation of the anomaly.

Dark Matter is the same in many ways. First, like I have said here many times, the name Dark Matter is a placeholder showing that we dont have the slightest clue as to what it is. Now we call it cold, or warm or hot dark matter, because we have an idea from its behavoir as to the internal energy of whatever DM is. The thing is tho, that we still dont have the slightest clue what DM actually is. So, to say "predictions from computer models fail to live up to the realty provided by observations" is kinda pointless. We dont know how good the model is, because to some degree the model is a guess as to all the properties and distribution of DM.

The thing is, even if we do get a pretty good idea of what DM is, that dosent mean the theory is perfect. Going back to the GR example, we still havent detected gravity waves directly. Should the current or next generation of detectors not find any, then GR will have a problem. Either the calculated intensity of the waves is off, or mayber there is something that damps them in space, or maybe our understanding of the experiment is bad. That doesnt mean we toss GR by the wayside any more than we tossed Newtonian dynamics. It only means that a new theory will need to be made that explains things better.

What's the question here? Whether it's useful or appropriate for scientists to disprove (falsify) a hypothesis without putting forward or pointing out the existence of a better hypothesis?

Most scientists realize this, but many people seeking to invoke falsification don’t. While it’s a useful concept, falsification can be easily abused because all science is inherently false or incomplete. Similarly, complex topics with many lines of evidence that all lead to the same conclusion like evolution cannot be falsified by any single test because even if something major failed the other lines would be more then enough to allow us to reach the same conclusion.

What falsification is meant to protect against is hypothesis that can only ever give “it’s true” for an answer. For example in another recent thread rtomes came up with thy hypothesis that the last 150 year of climate data can be explained by 4 overlapping climate cycles, but he gave no independent evidence for choosing those cycles and simply searched until he found 4 that gave him the match he wanted.

The problem here is that you can always find 4 overlapping sine waves that will match almost any climate pattern. There was never a chance that he would not be able to produce something that worked so the fact he did meant nothing. If it is never possible for a hypothesis to fail, then it’s success means nothing, that’s what falsification is about.

It's a general question: what - in your opinion - is the role of 'falsification' in modern science?

lomiller1, for example, has pointed out that it has merit as a criterion for deciding whether a hypothesis is within the scope of science or not (if the hypothesis is not, even potentially, falsifiable, then it can't be science).

korjik pointed out that a theory being falsified by a single experimental result (even if independently verified) doesn't mean that theory is necessarily abandoned; he also highlighted a very interesting aspect: domain of applicability (Newtonian gravity is just fine for a wide range of circumstances and applications, for example, despite its being comprehensively falsified in other domains).

In principle I agree with that, but it wasn't the original setting that there was centuries old theory which pretty much everyone knows has been falsified ages ago. It was about current theory that is generally considered as not yet falsified.

To say that "Y is falsified by set of observations X" in my opinion is very useful when Y is generally considered as not yet falsified. However, Nereid's attitude in this matter seems to be something like we should only talk about observation set X as a problem for Y if we already know how to fix Y so that X is not a problem for it anymore. That's somewhat similar to the situation where I would take my bicycle to repair shop but I wouldn't be allowed to tell them what's the problem if I myself wouldn't know how to fix it.

I'm not much of a fan of this current trend of going from "truth" to "truth", i.e. we consider one theory to be "truth" and we only abandon that theory if we have another theory which we can start considering as "truth". I would be quite happy to be officially in "we don't know" state, and just look for more observations to quide us, and perhaps do some low level hypothesizing.

__________________
"Stupidity gets denser in a crowd" - Old Finnish saying. [My website] [Nimblebrain forums]

This is an interesting, and I suspect, common, perspective; namely, that:

a) if something has been 'around' (in textbooks, etc) for a long time ("centuries"), that somehow makes it more immune to falsification;

b) if the something is, or could be, called a "theory"*, it gets to hang on longer in the face of (good, independently verified) observations that falsify it

c) 'falsification' cannot be applied retroactively, it can only be used for things within the last generation or so (~30 years).

In the context of "collisionless CDM", korjik's and lomiller1's remarks align quite nicely with yours Ari Jokimaki: "collisionless CDM" is neither vegetable nor mineral (there are, no doubt, whole classes of potentially falsifiable hypotheses that could be called "collisionless CDM", and the only 'theory' that comes even close isn't even called such ... LCDM cosmological models contain 'non-baryonic CDM', but its properties are extremely poorly constrained at the galaxy and kpc level, and below).

It seems that what I wrote is somewhat ambiguous, let me try again.

Let's replace 'falsify' with 'rule out', and add some riders concerning quantitative measures of how inconsistent a hypothesis is with (good, independently verified) observations. In this case {observations X} can rule out {hypothesis Y}, at the n sigma level (or something similar).

All well and good.

But what do you do on Monday, when you go to the lab? Y is dead, how do you go about formulating another falsifiable hypothesis? After all, it doesn't make much sense to ask for a million seconds of HST time if you have no idea where you're going to point it, what instruments you're going to use, and so on (let alone how you will use the data from those million seconds of observing to test something).

Back to beta decay: what status should we give the 'law' of conservation of energy, between the first discovery of 'missing energy' in such decays and 1957?

Or, more up to date, between Davis' first solar neutrino results and the discovery of neutrino oscillations, what (if anything) could have been said to have been falsified (by Davis' results)? solar models? the whole of the Standard Model of particle physics? And after neutrino oscillations were discovered, did one or other theory/model/etc become 'unfalsified'?

After all, 'falsification' isn't a terribly useful concept if its scope of applicability is highly ambiguous (or, worse, arbitrary or undefinable), don't you think?

* or, even better in the case of Newtonian gravity, a "law"!

snipped by me

There was alot of waiting with baited breath between the neutrino being postulated and it being detected. That was a very clear cut situation of is conservation a law or not.

Neutrino oscillation is an even better example. The neutrino anomaly meant one of two things, either we didnt understand neutrinos, or we didnt understand fusion. If there were no neutrino oscillation, then the calculations that gave us stellar structure that matches mass to luminosity would be wrong. If you continue down the path, the calculations that give almost anything atomic would be wrong, but that wouldnt be your biggest problem. Your real problem would be that all your incorrect calculations matched up with experiment. Now you have to explain how so many people got the right solution with an incorrect calculation.

Very interesting... and here I thought dgruss despised Popper.

__________________
"All your bias are belong to us." Ara Pacis
"A witty saying proves nothing." Voltaire

I think what you are driving at here, and I quite agree, is that what science needs to do is actually quite a bit more sophisticated than the binary operation "falsify/confirm". Many valid points have already been made above, but perhaps what I might add that has not been directly stated is that the concept of "falsification" must be seen in a direct context of what the original purpose was of the theory being falsified. If the purpose of the idea of ESP was to be able to detect it reliably, and possibly control it for some use, then it has been falsified in the "dead" sense of the word. If its purpose was to provide a vocabulary whereby we could fruitfully begin to investigate what we still don't know about the mind, then it could have been completely successful despite not being technologically applied (that it also failed in the latter pursuit is more the issue at hand, in the case of ESP). The point is, science is a process, not a destination, and models should be viewed in that same light.

This is never more true that for CDM-- what is the purpose of the CDM model? To say it is intended as a way of changing our view of our universe, as though it could transport our minds instantly into a position of knowledge, would be to substantially oversell its current place in astronomy. Really, it is intended as a way of answering the question "if we had to characterize all the shortcomings of our current view in the smallest possible space, what concept could fill that space". It's not time to change our view of the universe, it's time to recognize why we will still need changes in that view, and what we need is help finding what it is we have yet to discover that will change our view of the universe.

CDM is more like a set of instructions for following that course, than it is like a specific entity. As such, it cannot be criticized on the grounds that it does not answer all questions and has not been directly detected-- as long as it continues to make sense of many observations and motivate new ones, as long as it continues to give us a good perspective on what are the problems that don't make sense with other observations, and as long as it continues to guide our ideas of what types of substances we might look for to discover it, then it is serving its purpose admirably. We will know that CDM is dead only when we go a decade with no new mysterious observational results that it lends any insight into, and no new ideas of what experiments might detect it. That's not quite the same as "falsifying" a theory, it's closer to recognizing when a theory has become "dead weight". None of the criticisms leveled by dgruss23 ascend to that level, that I can see.

In summary, ESP is dead weight, ID is dead weight, and ufology is dead weight, not because any of these have been falsified (that would be impossible), but rather because none of them have proven useful for anything but getting books sold. Newton's gravity, on the other hand, has been "falsified", yet it is not dead weight, because it continues to be useful and we know exactly when we can use it and when we cannot. In short, it accomplishes a scientific purpose, while those others do not. For the time being, CDM also accomplishes its purpose. That may not always be so, but at the moment dgruss23's objections are premature (not completely invalid, just premature). I say this not to turn the thread into a discussion about CDM, as I know that is not Nereid's purpose, but rather to draw out the difference between what is "the truth" that could be "falsified" and what is merely productive-- and when we should or should not abandon hope of same.

This would be a perfect example of what I'm talking about. One need only imagine that the neutrino cross section was 1000 times smaller than it is. Perhaps that could be, perhaps some tinkering could preserve the universe as we know it but the neutrino would still not be detected today. Would we then say that conservation of energy was "falsified"? No, I don't think so-- not as long as the concept was still useful for suggesting ways we might look for missing particles, or for investigating the boundaries of when it applied and when it didn't. To be useful, a chainsaw doesn't have to be able to chop onions, it merely has to come with instructions about what it is intended to do.

__________________
Physics doesn't predict the future, it predicts the past that hasn't happened yet.

There are two kinds of delusions, the obvious kind that clearly don't work, and the insidious kind that clearly do.

No, what I said was "centuries old theory which pretty much everyone knows has been falsified ages ago". The emphasized part of my comment was my point, not the part about centuries old which I said only because Korjik's example just happened to be like that.

I disagree with this and with your c) too.

Even if Y is "dead", you still have all the observations left. Observations can guide you even if you don't have pre-existing hypothesis.

And with the thing I'm suggesting, we wouldn't concentrate so much on going from falsifiable hypothesis to falsifiable hypothesis, we would concentrate more on making observations than making theories. Falsifiability wouldn't be so important in that setting, but would only have a minor role.

__________________
"Stupidity gets denser in a crowd" - Old Finnish saying. [My website] [Nimblebrain forums]

But one must avoid falling into what might be called "google science", where you skip the organizational principles and just search over observational results. You could certainly do science that way-- observations would still be motivated by any absence in the database, and predictions would be possible via interpolation. But one of the key goals of science would be lacking-- the goal of achieving simplification and understanding of our complex reality.

This is what I get exercised about when I hear people (like Wolfram) argue that science has gotten too hard for our brains and needs to be turned over to computers. Science was always too hard for our brains-- the whole goal was to find ways to make it not too hard for our brains. That was always the toughest challenge in science, was Newton's genius, and continues to be our charge today-- even if it's not easy.

But Ari's point does raise an interesting point about falsification, which connects with the whole issue of what a theory is for. We can't decide if a theory is "alive" or "dead" until we recognize its purpose, and the purpose is often not to achieve arbitrarily accurate predictions-- it might be to help us understand something at a useful but not unlimited level of precision. In other words, it doesn't need to be "the last word", and so "falsification" is not cut-and-dried.

__________________
Physics doesn't predict the future, it predicts the past that hasn't happened yet.

There are two kinds of delusions, the obvious kind that clearly don't work, and the insidious kind that clearly do.

From what's in this thread, it seems that Popper's "naive falsificationism" is neither an accurate description of what happens in science nor a useful guide to how to actually do science.

The last few posts, by Ken G and Ari Jokimaki, have started to look at just what is the engine of science ... the role of theory for example, of unfocussed observation, purpose, and (goals). And for these areas, falsification(ism) is a minor player, at best.

This thread's pretty much done then?

I'll just clarify that I'm not suggesting that we just make random observations. My suggestion is basically just a shift of point of view; currently in science (at least in astronomy) there is strong emphasis on theories, and I would prefer more observation driven process.

I understand your point of view, but I disagree with it somewhat. I don't think we would lack the goal you are talking about, we would just get there very slowly, and in smaller steps. Currently we are trying to create set of theories that explain everything, then put all our resources studying that set of theories, and if that fails we do the same for the next set of theories. In my opinion, we are all the time dramatically wrong because we try to take it all the way (at least sort of) at once. If we would do it my way, we would be all the time quite correct, but we wouldn't "know" that much.

Mind you, I haven't really thought this through, and I'm also quite outsider when it comes to making science, so these are just idle thoughts I'm airing here, and there might be some traps in my suggestion I haven't realized yet.

__________________
"Stupidity gets denser in a crowd" - Old Finnish saying. [My website] [Nimblebrain forums]

That's a valid point. Indeed, my all-time favorite theorist was Feynman, and along with being a brilliant organizer and a profound thinker who always tried to get to "the core" of everything, he also always knew the observations that any successful theory had to predict. He never just said "this equation has to be right because it's in all the books", he would always cite an observation that the equation successfully explained. That shows he never fell so far "in love" with how things had to be theoretically that he lost sight of how they really are empirically.

I am all for a healthy sense of humility about what we "know" at any given time, as history has shown. Indeed, I would say the giant breakthrough of science came with Galileo and the simple recognition that we never know anything ahead of looking. Nature is its own authority, it need not conform to how we imagine it to be. But that just says why our theories should not be mistaken for anything other than our imaginings-- they still have an important place in science.

In my view the core goal of basic science is not to gain power over nature, as that exercise is a two-sided sword that brings in many more cultural and ethical issues, but rather it is simply to understand, to find what is simple and unifying and organizing among what is complex and chaotic and uncontrollable. When that is our goal, our "imaginings" are central, and whether or not they are "falsified" is more an issue of what the imaginings were supposed to achieve, moreso than if they are nature's real masters. Indeed, I think you are saying that nature is always the master of our theories, never the other way around, but still the theories are the purpose of the exercise-- because we already have nature.

__________________
Physics doesn't predict the future, it predicts the past that hasn't happened yet.

There are two kinds of delusions, the obvious kind that clearly don't work, and the insidious kind that clearly do.