In reading David Deutsch’s brilliant book, The Beginning of Infinity, I finally came across a couple of simple reasons why untestable theories in science are a dead-end and why the explanatory content of a theory matters. It’s very common for me to harp on about empiricism and evidence to friends and folk I debate on subjects like God, heaven, homeopathy, alternative medicine and other realms where science cannot speculate, or has to no avail. I’ve never, however, managed to condense such lectures into conversational fragments that didn’t make them hate me by the time I finished. For that reason alone, I’m glad I came across Deutsch’s book; for my argumentative arsenal has increased.
Let me start by asking a few questions:
Q1 – What is the single factor that science, pseudoscience, and non-science have in common? (This is not a trick question).
A1 – The answer is that they started thousands of years ago, with the same base of information, which is relevant to the conclusion at the end of this post.
Q2 – Now, what differentiates science/pseudoscience, and non-science?
A2 – Testability*
Put it that way, A2 is obvious. As Karl Popper wrote: empiricism is the demarcation point between science and non-science (the criterion of demarcation). In other words, the testability of a hypothesis will tell you if it can be improved by experience. And, if it can’t, there is nothing to rely upon except authority and the rejection of authority is what allowed the scientific method to come into being. This brings us to Deutsch’s first science nugget:
Deutsch’s 1st Science Nugget: an untestable theory cannot be improved upon by experience
There was nothing new in Deutsch’s nugget of wisdom that I didn’t know before, but its succinctness and comprehensibility are what struck me. No longer will I need to leap off onto fifteen minute tangents on why someone’s pet theory is wrong. It is as simple as this: if it can’t be tested, it can’t be improved by experience and is, therefore, far more likely to be wrong than right (which, in a Bayesian sense, may as well mean that it’s always wrong—in the sense that a rational person shouldn’t believe something a priori).
Another way to put it is this: when an untestable theory is set forth as an argument one has no good reason to believe it, as there are any number of equally untestable theories one could believe in. In such cases, there is nothing to resort to except authorities (which is to say ‘you’ll only have reason to believe this because he/she/I said so’—i.e., the way Natural News works!). Without testability, there are only arbitrary reasons for choosing between competing theories. Religion is the quintessential example of this phenomenon. How many religions are there? Hundreds. Christianity alone has something like 2,000 different sects. Furthermore, in Mankind’s short history, there have existed some 10,000 gods. According to Deutsch, the reason why is because the hypothesis that there exists a God or Gods is inherently untestable. Therefore, in the absence of rigorous empirical results, anybody’s pet reason for why this religion/sect/god is better than that religion/sect/god is as good as any other. (Actually, it is testable as Deutsch eventually avails, but only when the explanatory theory of the theory is bad, which I’ll get to soon).
An untestable theory (that is to say, a theory not able to be falsified) cannot be improved upon by experience, only by authority, and since an authority confers no explanatory power other than “because I said so,” it offers no real-life implications (other than those implications that it confers as a result of pretending it has real-life implications).
In Q2, I put pseudoscience on the same side as science when differentiating the demarcating factor between science and non-science. The difference between science/pseudoscience and non-science, as I wrote above, is testability. You can test science and find out one way or the other it’s validity, and you cannot do this in non-science. The reason that science and pseudoscience line up together is because you can test the pseudoscience. However, it’s staying power in the realm of its empirical demise is the difference between science and pseudoscience. That difference being their theories respective explanation. That is, whether a theory is a good explanation or bad is what further demarcates science and pseudoscience.
Duetsch defines a bad explanation as a theory that is easily variable while still explaining the same thing. For example, Deutsch uses the Greek myth for the explanation of winter: Persephone was forced into marriage with Hades and had to visit the Underworld to perform her womanly duty for 6 months every year. Persephone’s mother Demeter would become so sad upon her annual leave to the Underworld that she would blanket the known world in snow. That was the ancient Greek explanation for the winter. However, it is a bad explanation because it did not account for the seasons of the Southern Hemisphere, where it was summer alongside the Greek’s winter. Now, if replication was the sole factor in transforming a hypothesis into a scientific theory, an ancient Greek would need wait only year and after and upon winter’s recurring return, would have proved Demeter’s sadness to be the cause of winter. It follows then that replication is not the sole factor in determining the objectiveness of a scientific theory.
Imagine then, after our Greek scientist successfully proved his theory, that knowledge of the Earth’s sphericity was discovered, and the northern hemisphere’s winter corresponded to the southern hemisphere’s summer. Would they abandon their theory of Demeter’s winter? No, as Deutsch argues: the Greek scientist would have modified the theory to state that in Persephone’s annual absence, Demeter banishes the heat from her vicinity only, thereby explaining the South’s summer. Theory saved and Greek life marches on. Yet, what has changed in the explanatory content of the theory? Nothing! It essentially boils down to the same argument: the gods did it. By Deutsch’s definition, the theory is easily variable and is a bad explanation. A good explanation for the winter, on the other hand, looks something like this: the combination of the Earth’s 23.5 degree axial tilt, albedo, atmospheric components, and distance from the Sun along with the Sun’s thermonuclear transmutation of hydrogen into helium thereby releasing photons that, as a result of stellar convection, force the photons on a 10-million year journey to the Sun’s surface, upon which, breaking free, of the sun’s gravity dash out on an 8-minute sprint to the Earth all interact in such a way to produce regular annual variations in the average temperature on the surface of the Earth that humans call spring, autumn, winter, and summer. How easy is that theory to vary while still explaining winter? Nigh impossible. Remove the axial tilt component of the Earth, and our weather would be relatively unchanged year round (no explanation). Change the strong nuclear force—a key process of stellar transmutation—just 2% stronger and helium would not be produced in the sun’s core (in fact, no stars would exist at all because all the hydrogen in the universe would have transformed into bi-protons within minutes of the big bang resulting in a boring universe devoid of form). Again, no explanation. If any one of a million variables is changed, the end-result is entirely different. Therefore, it is a good explanation.
There we have the demarcation point between science and pseudoscience: The variability of the theory’s constituents should modify the theory’s explanatory power significantly. If it does not, be wary.
Deutsch’s 2nd Science Nugget: A hard to vary theory is more likely to be a good explanation than an easily-variable theory
Take the real-world example of homeopathy vs. medicine. They both started at ludicrously wrong theories of the human body. The conventional medicine of the time believed the body to consist of four humors: blood, yellow bile, black bile, and phlegm; while homeopathy believed that diseases can be cured by a ‘like cures like’ remedy that increases in a potentiality inversely correlated to the amount of remedy present. Despite medicine believing an equally silly thing at the time (though at least they could see those 4 things in the body), it has today come to the point of organ transplants, chemotherapy, antibiotics, vaccines, and a host of other lifesaving devices because it used the criterion of demarcation (testability) to remove the non-science components of then-medicine. By regularly testing new explanations for why certain things worked or did not work, and updating the relevant theories with better explanatory content. (I make it sound simple here, but in reality, it happened millions of times across billions of people with information being lost/rediscovered all the time. To top it all off, there were thousands of dead ends, duplicative works, hoaxes and the like. It was, by no means, an easy process, and it will continue for centuries more, so long as the core of science remains intact.)
Obversely, consider when homeopathic practitioners learned of the existence of atoms, their finitude, and pesky facts like Avogadro’s constant. They didn’t stop believing in the explanatory content of homeopathy (that substances which cause similar symptoms in large quantities in healthy people cure that same symptom in dilute quantities in sick people), they instead opted to invent a water memory that cannot be directly tested (just like our ancient Greek would’ve amended his theory of Demeter’s winter). Compared to the discovery of thalidomide’s effect on pregnant women, which was banned, with no excuses or fake hypothesises present, it shows the failure of homeopathy as a scientific theory.
In the end, the key difference between the competing theories is that even though they both started at horrible explanations, one strived not only for testability, but for good explanatory prowess, while the other did not. (Homeopathy has regressed into non-science because it embraced testability until it didn’t work, and can no longer be considered a pseudoscience by Deutsch’s definition.)
To summarize: real science revises its theories in light of new observations to arrive at better explanations, while pseudoscience does not. These are further differentiated from non-science by being testable, which separates theories that can be improved upon by experience from those that cannot.
The Beginning of Infinity is a fascinating escapade into science, philosophy, and epistemology and I heartily recommend to anyone so interested. Key points in the book include Deutsch railing against scientists for not taking their theories seriously, and ravaging pseudoscientists to shortsighted to see past their own bad explanations alike. And this post, by the way, is only material from the first chapter. I look forward to reading the rest of this book, interlocuting my own thoughts on his chapters in written posts like this.
Afterword: I am well-aware that I’ve defeated the purpose of why I like those two tiny tidbits of information by extrapolating them into a 1,800 word post! Some things never change. I’m okay with this.