Friday, August 12, 2011

Rationalism vs empiricism

Sean Carroll reasonably discusses the tension between "rationalism" and "empiricism".

These are two competing philosophical paradigms when it comes to the question what we need to acquire new knowledge. "Rationalism" (not to be confused with "rationality") says that we may find out the truth about the world "by pure thought"; "empiricism" says that we may only do so by constant "observations".

Before the Scientific Revolution, "rationalism" would be popular among the educated elites - various Greek philosophers and those who worshiped them later (including thinkers associated with the Church). "Empiricism" is mainly associated with the scientific revolutionaries themselves.

So there can't be any doubt that your humble correspondent agrees with Sean Carroll that the introduction of empiricism among the intellectual elites was totally necessary for the scientific method to flourish. The main scientific insights - at least those that were older than 100 years - were almost exclusively outcomes of the empirical approach to Nature.

At the same moment, I am not going to sling mud at all aspects of rationalism, either. The pagans were arguably an empirical culture - and they didn't have any real science, either. To develop science, we surely need to think a lot. We need to appreciate that things may sometimes be different than what they look at the surface. The more sophisticated science becomes, the more complicated logical and mathematical chains must be inserted in between the empirical observations on one side and our theories explaining them on the other side.

So I would probably agree with cosmologist Michael Turner's balanced approach - at least when it comes to physics. He said (at least during a Rutgers colloquium I once attended) that physics depends on the fine balance between the "hot and exciting conceptual speculations" on one side and the "cold and boring experimental data" on the other side. Whenever this balance is breached, physics degenerates either to philosophy or to botany.

In this co-existence of thinking and observations that has made physics so potent, observations have the ultimate power to decide what is right and what is wrong; thinking is needed to develop sensible candidate theories that can actually be tested. Both of these activities represent an important fraction of the scientific method. However, it's important to realize that we can no longer "establish" axioms by pure thought: the observations have gained an upper hand in this part of the process.

Historically, it turned out to be nearly impossible to develop science as we know it without making observations of the reality. After all, people had to do lots of them. When we present the history, most of those experiments look paramount. Sidney Coleman famously said that "not even thousands of philosophers meditating for thousands of years would manage to invent something as strange as quantum mechanics": some experiments were needed to convince us that Nature has very surprising properties at the fundamental level. However, I find it obvious that if the scientists' intelligence were higher, they could have avoided a vast majority of the experiments. A much smaller number of experiments could have been enough to converge to the image of the world as painted by the contemporary science.

Detailed, repetitive experiments may sometimes replace reasoning and vice versa. This is important to know in the fields in which experiments cost billions of dollars. A gang of theorists may often do some progress that may compensate at least a part of the experimental work. However, these theorists must still be scientists. They must realize that they're organizing empirical data. You can't replace experimental physicists by philosophers.

To make progress, the scientists must be good at listening to Nature via experiments and observations; and they must also be good at maths and the intuition needed to develop their theories in sensible directions.

Lee Smolin's medieval dogmas

The 13th comment under Sean's article was written by Lee Smolin. It is a stunning greeting from the Dark Ages. On one hand, Smolin totally denounces empiricism, emphasizing that his dogmas are enough to look for the truth. So you might think that he is a rationalist - except that his "rational" reasoning is totally irrational. It's just amazing how this guy sucks both in the empirical and theoretical approach to the real world.

He writes:
Dear Sean,

Just to clarify, I used the principle of sufficient reason (PSR) in the paper you cite and others to argue that any successful unification of physics must be background independent.
Well, this is just one example of the fallacy that Sean was trying to point out. Smolin starts with a dogma - structurally analogous to geocentrism, creationism, or any other dogma that people held in the past (whether it's analogous as far as validity goes remains an open question) - and he "demands" that Nature and scientists obey the dogma. He's not capable of even imagining that his dogma is wrong. And make no doubts about it, it is completely wrong.
This is not “against string theory” as many experts would agree that any precise, non-perturbative formulation of string theory must be background independent.
No real string theory experts could agree with this sentence because it is complete rubbish. Both AdS/CFT and Matrix theory are totally precise and non-perturbative - yet both of them are linked to specific backgrounds. More precisely, I mean specific asymptotic shape of the spacetime or, equivalently, a specific superselection sector of the Hilbert space of string theory. So we know that it can't be true that "any" description of string theory has the property Smolin mentioned. On the other hand, we may be sure that Smolin doesn't have a clue about string theory, not even at the most superficial popular level.

While both of the descriptions - AdS/CFT and Matrix theory - may describe arbitrary phenomena that appear in the bulk of the spacetime, the descriptions can't be used to change the asymptotic structure of the spacetime i.e. to switch to another superselection sector. Using an operational language, one would need to add an infinite amount of energy to the spacetime and - theoretically speaking - such an infinite energy would qualitatively change the description of the physical phenomena.

We can prove that string theory obeys the equivalence principle and that the diffeomorphism symmetry is exact in all effective descriptions of the stringy physics that contain the metric tensor. We can prove that in any known description of string theory - e.g. perturbative string theory - that allows one to start with different backgrounds (connected on the configuration space) or starting points, the resulting theories are equivalent (one background is a condensate of particles upon another background and vice versa) - in this sense, they are background independent. So physically speaking, we know for sure that physics of string theory satisfies all conditions brought by general relativity that it should satisfy. As long as you mean something physical by "background independence", we know that string theory satisfies it.

One could also dream about a description of string theory that describes all possible superselection sectors - all spacetimes, regardless of their asymptotic structure - using a common language. It's great. Some of us have looked or keep on looking. However, there is no scientific necessity that such a "universal and fully transparent" description exists at all. It is not guaranteed by the logical consistency and it is not guaranteed by the empirical data. It is just a belief - in fact, a very similar type of belief as a belief in God (because God is who may see the whole huge configuration space of string theory in its entirety, from a distance, while everyone else has to look at it from a place or region inside it).

And in science, beliefs simply have a much lower priority than the hard data. And you can't promote dogmas to something better by giving them fancier names, either. In the words of Sean Carroll,
I just don’t think the notion of “ontological necessity” is in any way useful. I can imagine many different worlds, with completely non-overlapping sets of characteristics. I see no difference between someone claiming that a concept is “ontologically necessary” and someone admitting that they can’t imagine a world without that concept. The correct prescription is to work to improve one’s powers of imagination.
This is the basic and most universal point that Lee Smolin completely misunderstands. That's why he still has no idea what science actually is, despite decades of life in which he tried to pretend that he has something to do with science. He doesn't.

By the way, the claim that Smolin's mumbo-jumbo isn't meant to be "against string theory" is surely amusing for those who know that the subtitle of his latest book says "the rise of string theory, the fall of a science, and what comes next". ;-)
The lack of such a formulation is a persistent weakness of string theory, but it may be one that can be overcome. I have certainly put a lot of effort into trying.
It may be weakness from a religious viewpoint but it is surely not a weakness from a physical perspective. This property of string theory as we know it today only contradicts an empirically unjustified aesthetic or philosophical dogma. But aesthetic or philosophical dogmas can't be treated as facts. They're pretty much irrelevant in science.

By the way, Smolin couldn't have possibly helped to solve any issue in string theory because he has no idea what string theory is.
In fact, most approaches to quantum gravity are as background independent as general relativity.
Physically, string theory is actually the only known quantum approach to gravity which is as background independent as general relativity. There is actually no other candidate quantum gravity theory that admits any backgrounds resembling those we know from general relativity. But let's ignore the wrongness of all Smolin's technical statements. Even if they were right, the very logic he uses is totally illogical.

What Smolin is incapable of understanding is that this property of the "most approaches to quantum gravity" - namely Smolin's interpretation of "background independence" - is demonstrably a part of the reason why they're worthless sewage of intellectual masturbation.

Imagine that you have a theory QG01. Then you have theories QG02-QG99. The theories QG02-QG99 satisfy a certain condition, C2, but they don't lead to any predictions that agree with the observations. On the other hand, QG01 that doesn't satisfy the condition C2 leads to an explosion of accurate science that qualitatively and quantitatively - and sometimes extremely precisely - agrees with many types of observations.

In science, this situation means that QG02-QG99 are disfavored, and together with them, their shared properties including C2 that are likely to be linked to the failure of the theories. The acquired evidence would surely lead a rational scientist to shrink the time he dedicates to QG02-QG99 and to focus on QG01. An irrational person won't be affected by any data or results: he will still spend 99% of his time with nonsense so that he will never have a remote chance to learn QG01 or anything that isn't nonsense.

There simply exists a lot of evidence that a valid theory of quantum gravity shouldn't admit an easy universal, background-independent description. The disagreement between the properties of the technically successful theory on one side and Lee Smolin's dogmas on the other side means that the latter are wrong. The science's "failure" to obey Smolin's dogma doesn't mean, as he seems to think, that science is wrong. It means that Smolin is wrong. The disagreement is actually Smolin's failure to agree with scientific facts.

After 100+ attempts to write a paper that makes any sense, Smolin is still waiting for the first one. Now, this means nothing for Lee Smolin. He may always write the 101st paper that doesn't make any sense, and so on. But for rational people, the continuing failure of Smolin - and all the people who share certain elementary assumptions with him - to write a single meaningful article in which something actually works even though it wasn't adjusted to work is a very powerful package of sociological experimental data.

It strongly suggests that something is seriously wrong with their whole way of thinking (or, more precisely, the lack of it). Smolins, Rovellis, and dozens of other alternative physicists - regardless of their being in a "majority" and "consensus" that physics should be a non-empirical science, and Smolin emphasizes this "consensus" - keep on failing to find any meaningful physical insights not just despite their would-be beautiful dogmas, but because of them. Valuable science simply can't be done if one constrains himself by tons of arbitrary assumptions that can't be rigorously mathematically or empirically justified. If you just guess them, they're pretty much guaranteed to be wrong.

Principle of sufficient reason

Sean also rightfully humiliates the "principle of sufficient reason". It's a philosophical assumption claiming that everything that happens and everything that exists must have a cause. All the data must be explained by something.

But this approach to knowledge is much like the five-year-old's approach. The kid may always ask "why". He will ask "why" after any other extra explanation you give him. Now, it's much easier to ask "why" all the time than to find explanations that actually make sense.

More importantly, it's damn obvious that at some point, asking "why" is no longer intelligent or productive. More precisely, it is less intelligent than asking many other "how" or "why" questions about totally different things. Every theory and every explanation must have some basic assumptions or axioms. The internal consistency and agreement with the empirical data decide which of the sets of assumptions and axioms are valuable and which of them are not.

In particular, the detailed quantitative and empirical data show that the assumptions underlying string theory are a golden trove while the assumptions underlying the remaining 98 theories are - if I may borrow Al Gore's vocabulary - just a pile of shit. You can't turn such a material in the pile into gold or platinum just by collecting 98 units of it.

In reality, we know very well that some things just don't have any reason. We know that classical determinism doesn't work in the real world; quantum mechanics shows that the outcomes of individual microscopic experiments are random. In the same way, there have been lots of historical coincidences that decided about the properties of the Earth, its orbit, and millions of other things that influence us today. Whether the masses and couplings of elementary particles can be "pretty uniquely explained" or are "nearly random" remains an open question.

Some of these properties could be given a more detailed explanation in the future than they have today. But many of them won't.

Many people love to say that "there are no bad questions". Many people have been led by the populists to believe that whenever they ask any question, they are very wise. But questions are not created equal. Some of them are better than others. And some of them are loaded: they may end with a question mark but they actually say much more about the opinions of the person who asks. If you reduce a set theory to 3 basic axioms and you will spend years by asking "why these axioms", you will probably get nowhere because they're not too good questions. It's unlikely that there's something more fundamental than these 3 axioms.

Theoretical physicists who want to make some impact must focus on certain sufficiently important questions. On the other hand, they must avoid excessively ambitious questions that have virtually no chance of being rationally addressed at the present time. The balance between the "audacity" and "doability" is another important consideration that decides about good science.

Bad physicists such as Smolin typically get stuck with some very narrow-minded assumptions and limited questions and then they spend the rest of their lives by convincing others - and themselves - that this focus is important. But because they do so, they're no longer doing science. What they're doing is marketing of shit. It smells and tastes so nicely, they tell everyone for 30 years. Needless to say, millions of laymen eat this shit for their dinner and they smack their lips, excited by the yummy delicatesse.

But the actual bulk of the scientific progress is happening elsewhere. It's not taking place in the realm of questions that were asked by Leibniz three centuries ago, questions that have seen no progress for three centuries, questions that remained (and, most likely, will remain) disconnected from anything that was happening in actual science, and questions that were only periodically abused by kibitzers to pretend that they're the new fashionable geniuses and saviors even though they didn't make a Planck length of progress themselves.

For a scientist, it is therefore extremely important to be able to live in the state of partial ignorance. Science at every moment must inevitably be ignorant about many questions. Some of the unknown answers have a chance to be found and some of them don't. A good scientist must become capable of figuring out quickly enough which of the questions are promising. A good scientist must be able to see that the principle of sufficient reason is wrong; to say the least, it is an extremely bad guide to decide about the things we may investigate because it essentially recommends us to mindlessly ask questions about random things and to get stuck in a random set of such questions. That's too bad.

The most important message of this text that tried to touch many issues is that the agreement with the empirically known features of the Universe is what ultimately chooses the right - and promising - theories. Philosophical dogmas can't ever replace the empirical data and dumb yet blinded bigots such as Smolin can't ever replace genuine scientists.

And that's the memo.


  1. Much of Deutsch's argument makes sense. His portrayal of early humanity is humbug. First, first many of the cave paintings are beyond the likely skills of the vast majority of untrained people living today, including Deutsch. Second, traditional (nonliterate) societies must store the entire culture in the members' heads. Thus, there is very little spare storage for knowledge not useful to everyday life. Even so, there sound archaeological evidence that the platonic solids were well understood as early as the Neolithic. Within an apparent few centuries of the invention of writing many of the rules of geomety we know of from Pythagoras were already current in the fertile crescent and probably elsewhere. Writing permits the accumulation of surplus knowledge and the recombinant study of diverse phenomena. Progress in science thus is the result of writing and an adequate population. It is stochastic rather than systematic.

  2. "many of the cave paintings are beyond the likely skills of the vast majority of untrained people living today, including Deutsch.

    Well, of course they are. But not as far beyond today's average person's likely skills as a Rembrandt, for exampl;, and that's the point I think.

    "archaeological evidence that the platonic solids were well understood as early as the Neolithic"

    I guess it really depends on what you mean by "well understood". Perhaps you mean that only that Platonic solids have been known since antiquity, and ornamented models of them can be found among the carved stone balls created by the late neolithic people of Scotland at least 1000 years before Plato (Atiyah and Sutcliffe 2003). Dice go back to the dawn of civilization with shapes that augured formal charting of Platonic solids.

    But this doesn't even come close to this:

    And this is on wiki -- anyone in the world with a computer can access this, not just a handful of folks.

    Anyway, I don't think Deutsch would disagree if what you mean is that early humanity was biologically capapble of understanding as much as we do