Soft Sciences Vs. Hard Sciences

[teo123]

You need to apologize if you want me to continue to respond to you.

OK, I am sorry if I misinterpreted your behavior. It's possible that I've learned the wrong lesson from my experience on the Internet forums.

Fancy ridiculous prism-kinda sky dome could refract light from sun and stars in pretty much whatever way you want, and then you have the overt hologram claims.

Then, just like with the Flat Earth Theory, the constellations would have different shapes depending on where we look at them from. That is, assuming that prism-kind sky obeys the laws of physics.

Then that shows disagreement in the field (lack of consensus) apparently.

It's more of an attempted falsification. It's a necessary part of science, right? The fact that it's done ensures there is a good reason to believe the consensus, that the apparent consensus is not just group-thinking. That's not done in pseudoscience.

Are you, or are you not, going to do a statistical analysis of published papers to prove your claims?

Why would I, when you basically agree now that economics is likely not more reliable than linguistics is?

OR... another ad hoc exception was fabricated.

OK, now you are just being ridiculous. An ad-hoc hypothesis is:
A: I have a dragon in my garage.
B: I would like to see it.
A: That won't work. It's invisible.
B: OK, then, let's spray it with paint.
A: That won't work either. It's incorporeal.
B: OK, let's then measure the temperature of the fire in his mouth.
A: That won't work either. That fire is heatless.
Ad-hoc hypotheses are, by definition, not testable, and they are made not to be testable.
This was like this:
A: I think the Havlik's Law is incorrect. Here is a relatively long list of words that appear to contradict it (a Proto-Slavic yer appears to be vocalized even though the Havlik's Law would predict it not to be).
B: Great! Let's analyze it... Those words don't really appear random. The vast majority of them are demonstrably ultimately from the Chakavian dialect of Croatian, and all or almost all of them have that unexpectedly vocalized yer in the first syllable. Instead of assuming the Havlik's Law is incorrect, we will have far better explanatory power if we assume it's correct, but that another sound change operated in the Chakavian dialect that made the yers in the first syllable vocalized regardless of the Havlik's law.
A: That would imply all the Chakavian words that had a yer in the first syllable in Proto-Slavic have the yer in the first syllable vocalized.
B: Are you aware of any counter-examples?
A: Not really.
B: Then maybe we have discovered another law. Let's publish it to see what other linguists who have studied it have to say about it.
It's like if you tried to predict the movements of charged balls. The Coulomb's Law will correctly predict it in most of the cases. However, once they are moving too fast, there will be apparent exceptions to it. And the magnetic force laws capable of explaining those apparent exceptions are in no way ad-hoc hypotheses.
Now, the special relativity is capable of explaining that the magnetic forces exist for the exactly same reason electrical forces exist. In linguistics, there is no such theoretical framework yet.

I just do not hold the kind of certainty as I would hold for chemistry or physics.

I mostly agree with that. Basic physics is without a doubt more certain than basic linguistics is.
You can test the Torricelli's law with the equipment you have in your kitchen, and it's very hard to get the data wrong. You can also test the Havlik's Law yourself, however, you can't be as certain that the data you are dealing with isn't wrong.
Now, is the modern physics more certain than basic linguistics is? I don't really think so. Doing proper experiments about the subatomic particles is very hard, both because you need to have a complicated machinery for that, and because the only laws we have about how they move are probabilistic. So, why is it that we should believe that the physicists are likely to get things right? Maybe we should say that basic linguistics is about as certain as advanced physics is.

You should tentatively stick with the consensus NOT just until you decide there's not a lot of good evidence for it, but until you have an actual alternative with BETTER evidence.

Like when I thought the horizon appearing to be rising as I climb is better evidence for the Earth being flat than the ships appearing to sink as they approach the horizon is for the Earth being round?

That's true, but consensus doesn't have to be that precise.

Consensus that isn't precise can be very misleading. Like, almost all the climate scientists agree that global warming is real and at least partly caused by humans. But, if people take it to mean all of the climate scientists agree global warming is more dangerous than a tyrannical government (as many people appear to), that's pretty dangerous.

Economics is also more politically charged than linguistics, which would be expected to motivate higher levels of disagreement.

So, you basically agree with me now?

Why do you think DNA analysis couldn't have predicted that?

How could it? Languages move way faster than genes do. If I get married, my wife and I can simply decide to speak English to our children despite both of us being Croatians, the genes can't stop us.
Also, there are some 1st and 2nd century CE sources that mention Croatians as a Scythian tribe. Croatians gave up their language in favor of a Slavic dialect, which was a more prestigious language where they ended up being. Today, Croatian is a Slavic language with few, if any, loan-words from the language of the Scythian (Indo-Iranian) Croatians.
Similarly, the language of the ancient Bulgarians was a Turkic language, yet the language of modern Bulgarians is a Slavic language with few, if any, loanwords from the Turkic Bulgarian.
What do you think happened to the ancient languages of modern-day Romania? That Romans killed all men and raped all the wives so that people would start speaking Latin there? People gradually dropped their languages in favor of the distantly related but culturally more prestigious Latin.

Why can't you just be agnostic to varying degrees?

Why aren't you an agnostic about it being necessary to eat meat?

That assumes a world without those things in some form is magically possible. There's no reason to believe it is.

If you claim something is necessary, the burden of proof is on you. Especially if you want to force it onto other people.
Besides, isn't the fact that Somalia was an anarchy for decades without anything terrible happening (if anything, it got better), and that Ireland was an anarchy for more than a century without anything terrible happening enough reason to think it's possible?

We need to push government to use more science and less dogma.

Which, if it's impossible for government to use science because of the way science works, is going to be very counter-productive.

which is that a new one pops up that's usually (but not always) worse.

Sounds to me much more like what's going on in Venezuela (where there is a very powerful government controlling everything) than what would be going on in an actual anarchy.

[brimstoneSalad]

Then, just like with the Flat Earth Theory, the constellations would have different shapes depending on where we look at them from. That is, assuming that prism-kind sky obeys the laws of physics.

The stars could be very far away. Or the stars could be an uncountable number of tiny lasers that each serve one person. There's also a way to make a screen where each dot has a complex prism on it that sends light out in different directions.
There are a few ways to create an apparently unmoving image with a near-by surface. They're all pretty high-tech, but within the scope of physics.

It's more of an attempted falsification. It's a necessary part of science, right? The fact that it's done ensures there is a good reason to believe the consensus, that the apparent consensus is not just group-thinking. That's not done in pseudoscience.

How was that claim shown to be false? Or did others simply reject it for no good reason? (Perhaps as they reject your own theory)
Rejection of theories in softer sciences sometimes have more to do with group thinking than actual evidence.

OR... another ad hoc exception was fabricated.

OK, now you are just being ridiculous. An ad-hoc hypothesis is:
A: I have a dragon in my garage.
B: I would like to see it.
A: That won't work. It's invisible.[...]
This was like this:
A: I think the Havlik's Law is incorrect. Here is a relatively long list of words that appear to contradict it (a Proto-Slavic yer appears to be vocalized even though the Havlik's Law would predict it not to be).
B: Great! Let's analyze it... Those words don't really appear random. The vast majority of them are demonstrably ultimately from the Chakavian dialect of Croatian, and all or almost all of them have that unexpectedly vocalized yer in the first syllable. Instead of assuming the Havlik's Law is incorrect, we will have far better explanatory power if we assume it's correct, but that another sound change operated in the Chakavian dialect that made the yers in the first syllable vocalized regardless of the Havlik's law.
A: That would imply all the Chakavian words that had a yer in the first syllable in Proto-Slavic have the yer in the first syllable vocalized.
B: Are you aware of any counter-examples?
A: Not really.[...]

This is fundamentally the same thing. You're working backwards from observations and making exceptions to the original claim.

1. There's a dragon/Havlik's Law is correct
2. Here's an observation that doesn't seem to fit that (can't see it or whatever)
3. Well that must be because of X correlation, so that's an exception that changes our expected observations of 1 such that 2 doesn't disprove it.

You aren't helping your case here, you've basically just written a how-to of pseudoscience...

It's like if you tried to predict the movements of charged balls. The Coulomb's Law will correctly predict it in most of the cases. However, once they are moving too fast, there will be apparent exceptions to it.

That's not just an exception at "too fast", it's wrong for any relative motion because it doesn't account for the effects of relativity. This isn't ad-hoc, it's experimentally verified in too many ways to count, and we know precisely why that is.

Relativity upended physics. Everything dealing with relative motion was just wrong, and at a fundamental level. Not very wrong at low speeds, but still wrong. Coulomb's can only be used at low speeds as an approximation, it's not actually true. It's like assuming a cow is a sphere.

We also don't consider Newtonian physics to be a true representation of reality, just a useful approximation.

Now, the special relativity is capable of explaining that the magnetic forces exist for the exactly same reason electrical forces exist. In linguistics, there is no such theoretical framework yet.

That's what's missing, and that's what makes it ad hoc. Both the theoretical framework and the experimental verification (not just working backward).

Now, is the modern physics more certain than basic linguistics is?

If you're talking something like string theory, maybe not, but I think I've said enough about how that's more of a model that doesn't really make verifiable predictions.

Doing proper experiments about the subatomic particles is very hard, both because you need to have a complicated machinery for that, and because the only laws we have about how they move are probabilistic.

You're mixed up there, confusing wave form probability with the colloquial sense of probability as unknowns (unknown unknowns too); the former is a rigorous mathematical system with equations that explain it very clearly (uncertainty principle), the later is immeasurable; you can't figure out the probability of something mathematically in that colloquial sense.
You can actually get very reliable results with a clear p-value from a collider, so that's a very very hard science, even if it's also hard to do (and expensive).

So, why is it that we should believe that the physicists are likely to get things right?

See above.

Maybe we should say that basic linguistics is about as certain as advanced physics is.

No. Although you could compare it to string theory if you wanted.

You should tentatively stick with the consensus NOT just until you decide there's not a lot of good evidence for it, but until you have an actual alternative with BETTER evidence.

Like when I thought the horizon appearing to be rising as I climb is better evidence for the Earth being flat than the ships appearing to sink as they approach the horizon is for the Earth being round?

Only if you had made your model of that and shown it to be true by experiment.
A hypothesis isn't a good reason to reject consensus.

I'll try to answer the rest later. Might be a while.

[teo123]

There are a few ways to create an apparently unmoving image with a near-by surface. They're all pretty high-tech, but within the scope of physics.

OK, maybe. I don't have the knowledge needed to evaluate that statement.
Even then, the theory that says the sky is like that wouldn't really be a coherent theory (the way Bradley defines coherence), it would at best be merely a consistent theory.

How was that claim shown to be false?

Simply, the Havlik's Law makes unambiguous predictions about which yer (short 'i' or short 'u') in some Proto-Slavic word will become vocalized in modern Slavic languages and which one will disappear. Thankfully, it's a lot easier to test than "'Issa' meant 'health-giving spring' in Illyrian." or "'Issa' meant 'island' in Pelasgian.".
It's the only law that determines what happens to the Proto-Slavic yers in, as far as I am aware, all Slavic languages and dialects except the Chakavian dialect of Croatian. In the Chakavian dialect of Croatian, the Havlik's law appears to be invalid... unless you take into account that there was another sound law operating in that dialect that made all the yers in the first syllables vocalized.

Perhaps as they reject your own theory

Just... no.
Like I've said, my theory was rejected primarily because I, back then, hadn't done my homework of specifying what sound laws I think operated when an Illyrian word was borrowed into Old Croatian. That has to be specified, unless I assume Old Croatian and Illyrian somehow happened to have the same phonology. If I don't specify that, my theory only makes predictions about the ancient Croatian toponyms, and it predicts nothing about the modern Croatian toponyms, and it's therefore a lot harder to falsify.

Rejection of theories in softer sciences sometimes have more to do with group thinking than actual evidence.

This can hardly ever be stated for certain. Flat-Earthers often cite group thinking as the reason why astronomers reject their supposed evidence.

This is fundamentally the same thing. You're working backwards from observations and making exceptions to the original claim.

No, ad-hoc hypotheses are unfalsifiable assertions about why some way of testing a claim won't work. Claim "All the Proto-Slavic yers in the first syllable became vocalized (turned into 'a' rather than disappearing) in the Chakavian dialect of Croatian." is not unfalsifiable.
I am really confused about how you think science works. The heart of science is that theories are being tested and modified if they don't make correct predictions, and rejected if they can't be modified to be consistent with the observations while staying falsifiable.

That's not just an exception at "too fast", it's wrong for any relative motion because it doesn't account for the effects of relativity.

I meant, the discrepancies won't be visible unless the measurements are very precise. The words in the languages can be thought of as some sort of discrete measurements of the sound laws that operate or have operated in the language or the dialect that word is from. A single measurement can't tell us about all the laws that operated, since any particular word will not be of the shape necessary to be affected by all the laws that operate of have operated in a language (and, even if it were, there would be many possible explanations for what happened to it so that it sounds the way it sounds). But, if we have many words, we can discern those laws.

That's what's missing, and that's what makes it ad hoc.

So, the theory of electromagnetism was ad-hoc before the special relativity was discovered?

You can actually get very reliable results with a clear p-value from a collider, so that's a very very hard science, even if it's also hard to do (and expensive).

Fine, maybe you can. I don't have the knowledge needed to evaluate that claim. But, even so, that requires way more than the basic understanding of the laws. So, again, why should we trust physicists to be able to do that more than we should trust the linguists to be able to figure out the sound laws?

[brimstoneSalad]

Even then, the theory that says the sky is like that wouldn't really be a coherent theory (the way Bradley defines coherence), it would at best be merely a consistent theory.

Consistency is adequate. Why are you bringing up Coherentism? It's not something I'd generally recommend...

Simply, the Havlik's Law makes unambiguous predictions about which yer (short 'i' or short 'u') in some Proto-Slavic word will become vocalized in modern Slavic languages and which one will disappear. Thankfully, it's a lot easier to test than "'Issa' meant 'health-giving spring' in Illyrian." or "'Issa' meant 'island' in Pelasgian.".

Assuming we could figure out the chance of that happening by random drift, that would work if you had some mutually isolated islands where the people there haven't been contacted in a thousand years and we honestly have not been exposed to their language. Then a prediction could be tested by visiting each island one by one as multiple uncontaminated samples which have not biased the hypothesis.

Again, even assuming you work out the probability of it happening by chance if language drift is random, there are two problems with the "test" of Havlik's law:
1. Not enough samples. The p-value for this would be terrible. Sounds like 0.50 (a 50% chance of happening if it's random). You'd need at least five samples to get a decent p-value. How many different languages have been tested?
2. The samples themselves weren't isolated, and could have contaminated the hypothesis (making it ad hoc, because this may have been consciously or unconsciously known to some degree before the hypothesis was developed).

When your "test" is merely observing something again that already exists and that you may have already observed while or before forming your hypothesis, that's a big issue.

Is it possible that this is a credible test? Sure.
Again, I'm not at all saying that this isn't on the spectrum of science. Even the softest sciences have occasional tests that follow good methodology.
The point is that it's a spectrum, and these kinds of tests are more rare, usually have weaker p-values (if any), etc.

If I don't specify that, my theory only makes predictions about the ancient Croatian toponyms, and it predicts nothing about the modern Croatian toponyms, and it's therefore a lot harder to falsify.

You can't make predictions about something you already have knowledge of...

It's like I hypothesize that cilantro tastes like soap to humans, I test myself by eating some cilantro and deciding if it tastes like soap or not, and then I determine that it tastes like soap and consider my experiment a success. Never mind that I've had cilantro before thus biasing the hypothesis with the already known result, and that the sample of one is a poor representation of humans in general.

Now assuming you don't know any Chinese, you might be able to form a hypothesis about how phonemes develop in Chinese and *then* have a more credible test of it. There's the possibility that Chinese you've heard in movies but didn't understand unconsciously biased your hypothesis, but much less so than a language you've studied and are familiar with. It's unclear the degree to which humans are capable of subconsciously learning language and sound patterns by osmosis without paying attention, but I would never trust a hypothesis about Croatian from somebody who speaks it to be uncontaminated by foreknowledge.

This can hardly ever be stated for certain. Flat-Earthers often cite group thinking as the reason why astronomers reject their supposed evidence.

The reason that isn't credible is because astronomy is a harder science, you can't group think your way around obvious and irrefutable falsification. You can group think your way around unfalsifiable ad-hoc models.

As you move down the spectrum into softer sciences, the more stable group thinking is. As you move up into the hard sciences, they'd have to be completely insane or it'd have to be a conspiracy.

No, ad-hoc hypotheses are unfalsifiable assertions about why some way of testing a claim won't work.

It's about why that result of the test doesn't falsify it. There's nothing about assertions like that which have to be unfalsifiable. The added claim could very well be falsifiable *if* it were not amended again to skirt falsification.

An ad hoc hypothesis is about making it so a particular result of a test of the claim doesn't (or won't) invalidate it by amending the hypothesis (or creating it in order) to be immune to falsification from that test result. Then typically rinse and repeat for every objection.

It in effect amends the original hypothesis. That's what makes it ad-hoc, it refers to the hypothesis itself. It changes it to fit the details after the fact, or has been created after the fact.

Which is basically what arbitrary exceptions are.

"Well this law is true except for all of these exceptions where it isn't true"
That has no value, because we don't know if a new case is an exception or not. Basically, the law isn't true, it's just sometimes that sometimes happens and you really have no idea why.

Claim "All the Proto-Slavic yers in the first syllable became vocalized (turned into 'a' rather than disappearing) in the Chakavian dialect of Croatian." is not unfalsifiable.

That's more of an an observation than a hypothesis. But it isn't really falsifiable because all we know is that the *surviving* words that we know about with yers have that. It's induction.
Like "All of the apples I've seen are red, therefore all apples are red" except here there aren't any more apples to check.

A hypothesis, in order to really be valid, needs to be applicable to an as of yet unknown data set.

Again, like the uncontacted islands thing. Or discovering a scroll from a dead language or something.
It happens sometimes, but it's rare. That limits linguistics as a science.

The heart of science is that theories are being tested and modified if they don't make correct predictions, and rejected if they can't be modified to be consistent with the observations while staying falsifiable.

You can modify a theory, but if you have to do that then it starts back at square one and you need to test it again. The failure to re-test, but just modifying it with an exception that's consistent with past observations, is what makes what you described ad hoc.

But, if we have many words, we can discern those laws.

Again, you're working backwards: it's ad hoc.

So, the theory of electromagnetism was ad-hoc before the special relativity was discovered?

No, it was just wrong. Then it was replaced and retested. If it were not retested then it would have been ad hoc.
You start over completely with reaching statistical significance, you can't use the results for any of the tests you've already done. That means you have to find a new set of five languages you've never encountered before to test.

But, even so, that requires way more than the basic understanding of the laws. So, again, why should we trust physicists to be able to do that more than we should trust the linguists to be able to figure out the sound laws?

It requires specialized equipment... what they're doing with supercolliders isn't that complicated. They're reaching certain energy levels and looking at what particles come out of that based typically on their decay products and the energies of those products. They can run it *many* times, thus achieving statistical significance.

It's as if linguists could create isolated islands and put people on them, make a thousand years elapse, and then explore how those languages have changed. And if they could do this hundreds and thousands of times.

If linguists *could* do that, then linguistics might be a hard science like physics. But they can't.

Maybe once we can simulate human brains on computers we could do that, and then you'd have real statistically significant tests of linguistic hypotheses.

But like I said, despite being a soft science I DO assume linguists are correct because I have no good reason to believe otherwise.

[teo123]

I think I can address all or almost all of your points here just by addressing two of them...

You can't make predictions about something you already have knowledge of...

Therefore, Galileo was wrong to make predictions about how things move, just because he had a lot of experience with that?

Again, you're working backwards: it's ad hoc.

Therefore, nothing can ever be learned from history? Well, I guess that could be true for economics and political science, but I think linguistics is a little too simple science for that to be true for it.

[brimstoneSalad]

I think I can address all or almost all of your points here just by addressing two of them...

You can't make predictions about something you already have knowledge of...

Therefore, Galileo was wrong to make predictions about how things move, just because he had a lot of experience with that?

The human brain evolved to process language and learn unconsciously what "sounds right" or not. Processing the relative positions of countless stars and other bodies from sight isn't quite as plausible.

That said, in so far as the measurements and movements were known (and many of them were) and he worked backwards what he was doing WAS creating a model.

After you have a model you can start making theories to predict actual unknowns. The heliocentric model and gravity's role in orbit have since been confirmed with experiments that tested things that were not known to Gelileo, so they're very hard science at this point. But in his day it may not have been as rigorous. You'd have to ask a historian of science when it was really confirmed with new information and no longer just a model to explain things that were already known more eloquently.

Also, what made the heliocentric model superior to the geocentric one was that it was far simpler and mathematically elegant. The geocentric model was a mess as you probably know.
When you compare models with no actual empirical support, all things being equal, you prefer the simpler one.

What you're probably wondering:

Should lay people of the time have immediately abandoned geocentrism?

Probably not until some new evidence (and not just the model) were presented in favor of the heliocentric model.
This on the basis of it not being consensus yet (it took a little bit given the time, but it was consensus among astronomers long before the church accepted it), and the geocentric model coming first. Non-experts are not usually qualified to evaluate how simple or complex a model is as per Occam's razor. You just need to wait for the experts to catch up when that happens.
It's unfortunate that non-experts by necessity need to be a little behind the curve (thankfully it's no longer measured in decades), but that's to prevent people from upending the systems they're using for every compelling quack theory that comes along. It *IS* usually best to sit on these things for a bit until experts weigh in. You can see the same thing with health trends in pop culture, anti-vaxx (there was *once* a peer reviewed paper, since retracted), whatever.

Again, you're working backwards: it's ad hoc.

Therefore, nothing can ever be learned from history? Well, I guess that could be true for economics and political science, but I think linguistics is a little too simple science for that to be true for it.

...There's a good reason that history is not science.
It doesn't mean we can't *learn* anything from it, but it's not the kind of knowledge that has scientific precision or probability values.

It's like "Hey, last time something that seemed kind of like this happened it was bad, maybe we should tread carefully?"
It could be completely wrong because we haven't controlled for any variables, but it at least tells us what to look at.
It's a starting point to direct us to some things we might want to study. History is more like an anecdote, or collection thereof. As in medicine it's not direct evidence, but it can tell us what we should look into and study.

For example, the lack of successful communist governments in history doesn't tell us that communism doesn't work, but that we should be skeptical, particularly of certain aspects of implementation. We can look at what failed in the past, speculate on why, and then do actual contemporary studies to see if we can fix those problems.

[teo123]

The human brain evolved to process language and learn unconsciously what "sounds right" or not.

Are you aware of some evidence of that that I am not? I mean, simple machine learning appears to also be able to learn the grammar of human languages. Besides, even if we did somehow evolve to be able to quickly learn a language, we certainly didn't evolve to be able to describe it. If you ask a native speaker to describe some rules of the grammar, they will often describe it incorrectly. I know that, because it has often happened to me.

But in his day it may not have been as rigorous.

So, you basically admit that every science needs to start as what you call "soft science"? You know, like having the "laws" for which we don't immediately know where they apply and where they don't apply?
I mean, the electromagnetism basically evolved like this:
1. It's been observed that electrically charged bodies cause a force on each other.
2. That force has been described quantitatively using a relatively-simple Coulomb's law.
3. It's been shown that the Coulomb's law appears to be correct except when those charged bodies move very fast.
4. The Biot-Savart's law and the Lorentz's force law were formulated that, among other things, quantitatively described those apparent exceptions to the Coulomb's law.
5. The Special Relativity was discovered, and it turned out to be able to, among many other things, explain all those laws as in fact being a single law.
Likewise, the theory surrounding the Havlik's law evolves like this:
1. It's discovered that Slavic languages are related between each other and related to Latin and other Indo-European languages. Systematic study of the ancient inscriptions on the Slavic languages begins. It's been observed that the short 'i' and the short 'u' (the "yers") in ancient Slavic languages sometimes correspond to a single vowel in modern Slavic languages (different for each Slavic language, but consistent within that language) or disappear, but the rules for which yer gets vocalized and which disappears remain unclear.
2. The exact description of the rule those yers follow is formulated: it's the Havlik's law. Unlike for the Coulomb's law, which was very similar to the already-known Law of Gravity, the Havlik's law was rather unlike anything known to linguistics at that time.
3. The Havlik's law is shown to be correct except for some words coming from the then-little-studied Chakavian dialect of Croatian.
4. Another law is formulated to exactly describe those apparent exceptions to the Havlik's law.
---this is where we are now---
5. In the future: a unified theory is discovered that explains, among other things, why that first-syllable-exception is to be expected in some dialects where the Havlik's law operated.
You can know that something works without knowing how, and you can know how something works without knowing why (if science indeed ever fully answers that question, it's hard to tell whether science actually explains things or merely describes them). That's true for all science, not just social sciences.

There's a good reason that history is not science.

I've never heard anyone saying that before. History sometimes makes incredibly precise predictions. Like, it predicted how old the Narmer stone is with the precision of three centuries (and the chronology suggested by Champollion a century earlier was off by millenniums), which the radiometric dating confirmed.

It *IS* usually best to sit on these things for a bit until experts weigh in.

Except that, like I've said, for most issues, you don't have a way to estimate how many experts actually agree on something.

For example, the lack of successful communist governments in history doesn't tell us that communism doesn't work, but that we should be skeptical, particularly of certain aspects of implementation.

So, you admit that those statistics showing how historically too much social and economic freedom caused violence don't tell us much?

Also, why did you say you suspect sound laws to have bad p-values? I just don't see it. Like, on the top of my head, I can think of quite a few words in which English 't' corresponds to German 'z' in the beginning of a word, and I don't speak much German: two-zwei, ten-zehn, tooth-Zahn, tongue-Zunge, twig-Zweig, toe-Zeh, too-zu, tame-zahm... How do you mean it's likely for such a pattern to arise by chance?

[brimstoneSalad]

Are you aware of some evidence of that that I am not? I mean, simple machine learning appears to also be able to learn the grammar of human languages.

Of course they can, these systems operate on the same principles of evolution and human cognition.

Besides, even if we did somehow evolve to be able to quickly learn a language, we certainly didn't evolve to be able to describe it. If you ask a native speaker to describe some rules of the grammar, they will often describe it incorrectly. I know that, because it has often happened to me.

That's beyond the point. The point is that you may already have unconsciously absorbed the language patterns, so your assessment may just be a matter of working backward from the known patterns to ad hoc rules that describe those patterns.

So, you basically admit that every science needs to start as what you call "soft science"?

Of course scientific method has improved over time. I'm pretty sure I already said pretty clearly that a soft science can BECOME a hard science through innovation in a field.

You seem to think you have a "gotcha" or admission here, but I've always maintained this.

Linguistics could some day become a hard science, but as of now it is still a soft one.
However, it doesn't work like you think it does.

You know, like having the "laws" for which we don't immediately know where they apply and where they don't apply?
I mean, the electromagnetism basically evolved like this:

I was very clear that what makes these things hard science is experimental verification. Without that it IS just modeling.

Likewise, the theory surrounding the Havlik's law evolves like this:

No, not likewise. The lack of experimental verification is a currently insoluble problem.
Everything you described is ad hoc.

Let me give a clear example:

You have ten apples. One is green, the other nine are red.

Here's a model consistent with that observation:
90% of apples are red and 10% are green.

That's not science.

Now what would make it science is if you found 100 more apples you've never seen before and tested them to determine if they fit that model. That's experimental verification, and it's what soft sciences tend to lack.

What you described is even worse.

You have ten apples, all of them are red.
You say "100% of apples are red"
Then you find another one, and it's green. So you look for an ad hoc explanation to justify it as an exception. Turns out it has a split stem. So now you say:
"100% of apples are red, except apples with split stems; 100% of them are green"

Not science.

Again, you need experimental verification AFTER your hypothesis to make it conform to hard scientific methodology.

The rest of your comments seem to be under a similar misunderstanding. Let me know if you need me to address them.
I'll note: this just stands as proof that you don't understand anything about science, and as evidence that experience in a soft science does little to nothing to prepare somebody to understand science in general and particularly hard sciences. Ad hoc models are not the same as actual experimentation.

[teo123]

Look, @brimstoneSalad, I think I understand what you are saying, but that seems just so... wrong, you know... what you are saying appears to imply that almost all the ways of thinking and arguing we are using in a day-to-day life are very fallacious.
So, you are saying that it's practically impossible to apply the scientific method to languages and that therefore practically nothing about them can be really known, right? But if so, how come have you been able to learn English (and other languages you speak) so well? You are implying you couldn't have applied any method that would have actually increased the chances of your guesses about the English grammar being right... yet your guesses about the English grammar seem to be rather good.

[brimstoneSalad]

Look, @brimstoneSalad, I think I understand what you are saying, but that seems just so... wrong, you know... what you are saying appears to imply that almost all the ways of thinking and arguing we are using in a day-to-day life are very fallacious.

Well, yeah. Most people are drastically over confident or completely misunderstand the nature of knowledge.

Not sure why this would be a surprise.

So, you are saying that it's practically impossible to apply the scientific method to languages and that therefore practically nothing about them can be really known, right?

Certain parts of linguistics are pretty hazy.

In terms of contemporary descriptive linguistics -- like how grammar rules (the way people usually use words as observed) and such function today -- we can absolutely know them very well, but that's just a descriptive application. It's as precise as its measurement, and that's all. It doesn't form any predictive theories about how languages evolve. Those predictive theories are weaker due to inability (or limited ability) to verify them.

You can absolutely verify contemporary descriptive claims of usage just by polling people. That's what a usage panel is for.

But even though predictive claims about how languages evolve etc. are softer science, it's not completely without basis. Once in a blue moon you find a relic to confirm things.

But like I said before, regardless I still believe linguists when they tell me of the various laws etc. I accept these things as true because I have nothing else to compete with it and no real reason to doubt it.

To give perspective:

For example, the lack of successful communist governments in history doesn't tell us that communism doesn't work, but that we should be skeptical, particularly of certain aspects of implementation.

So, you admit that those statistics showing how historically too much social and economic freedom caused violence don't tell us much?

It's more that it doesn't give us a lot of confidence, but it tells us something. Without any other information available, a reasonable person accepts that as tentatively true.
Of course any contradicting hard evidence at all would completely override that.

It's like if you find something in biology and then chemistry tells you that's impossible, well then you go back to the drawing board and figure out what you did wrong.
If you find something in Chemistry and then physics tells you that's impossible, then again back to the drawing board; you messed something up.

We should accept what experts in a field tell us unless something else in a more rigorous, scientifically harder, field contradicts that.

So history does tell us anarchy isn't a good idea. It doesn't tell us with the amount of confidence that would lend the kind of certainty we have in physics, but it's what we should accept until other better evidence contradicts it.

But if so, how come have you been able to learn English (and other languages you speak) so well? You are implying you couldn't have applied any method that would have actually increased the chances of your guesses about the English grammar being right... yet your guesses about the English grammar seem to be rather good.

Again, that's a contemporary descriptive issue.

Non-scientific (or less scientific) fields can be very good at collecting descriptive data without very rigorous methodology and without developing any strong theories from that. The trouble is when they try to draw broader conclusions.

You poll a thousand people and ask them what "dog" means, and if 99.9% of them point at a furry mammal wagging its tail and barking, you have a pretty good inductive guess that's what it means in that population. Start using it that way and if nobody seems to get confused and you're probably safe.

But of course human language is imprecise. Like I said before it's hard to really prove that two people understand each other. Similar to the question of any qualia (does what looks red to me actually look blue to you, and we both just call it red?).

We accept these things because we don't really have a choice; there's not really an alternative. Radical skepticism doesn't get you anywhere.