It’s the end of the year, so this seems like a good time to talk about something grand, vague, and slightly hollow—though that has probably been true of earlier pieces too.
A common picture of science goes like this: observe the world, summarize experience, extract patterns, then apply those patterns back to reality. That is indeed how much scientific work operates. But historically, that route was not always regarded as rationalist at all. In the same way, the cliché that students in science and engineering are “better at logical thinking” usually comes from mixing together several different ideas without sorting them out.
It is worth separating reason, rationalism, science, scientific research, and logic, because they are often bundled together as if they were naturally the same thing.
Roughly speaking, reason stands opposite to feeling. Reason is grounded in logic; feeling is grounded in personal sensation. Between them there is something like understanding or cognition: knowledge that arises from sensory experience but has not yet been abstracted into full rational form. Most ongoing research is really focused on the move from sensation to knowledge. The next step—from knowledge to reason in the stronger sense—quickly becomes slippery, almost transcendent, difficult to state clearly.
There was also a long period in which rationalism and empiricism stood opposed to one another. Rationalists believed that starting from a few basic logical principles, one could deduce the structure of the world. In that view, already-given reason could generate knowledge. Unsurprisingly, people who thought this way were often mathematicians, and many of them had clearly spent time with Elements. Empiricists argued the opposite: knowledge comes from life in the real world, from sensory experience.
Seen from that angle, rationalists should reject information delivered by the senses. And in fact that austere attitude was especially beloved by serious religious thinkers. Many held that God revealed His existence and divinity through reason itself. Practical, sensory, worldly things were seen as inferior, even aesthetically degraded. Yet when people later began to articulate a philosophy behind science, the first thing they affirmed was almost the reverse: scientific knowledge must come from demonstration and verification, and what cannot be verified should not be studied.
The irony is that the methodology they used was still largely inherited from rationalist forms of argument. Science, from the beginning, was therefore a kind of hybrid creature: empirical in where it gathered material, rationalist in how it tried to organize and justify it.
The weakness of induction, and the compromise of falsification
Induction is not logically rigorous. No matter how many white swans you have seen, you can never rule out a black swan tomorrow. So knowledge extracted from observation is never complete.
One standard way to smooth over this problem is to lean on deduction, which is logically stricter. This is what is usually associated with falsification. I put forward a hypothesis, then let observational data test it, and at some confidence level I make a decision. It is important to note that science has always been advisory when it comes to decisions. It does not command; it recommends.
In that sense, the hypothesis testing used everywhere in scientific research is an embodiment of falsification, and therefore of a certain logical seriousness. But this, too, is a compromise. Hypotheses are endless. You can eliminate several, but you cannot exhaust the space of possible hypotheses. More importantly, you can only show that a hypothesis fails; you cannot conclusively establish that it is true. If taken strictly, that means you have ruled things out without necessarily generating new knowledge.
And yet in real research, hypothesis testing is almost always framed as a contest between the null hypothesis and an alternative. Once the null is rejected, people shift to the alternative, even though that move is not especially strong in pure logical terms. Still, statistical inference survives and remains central.
Why? Because choosing one statistical method rather than another is often more art than logic. It is common to see the same dataset produce different decisions under different methods. The source of the discrepancy is not always the data itself. It lies in decisions beyond science, in human beings, and human complexity introduces uncertainty as something built-in.
And decisions are historical. They cannot wait until logic has been fully satisfied.
Science is useful partly because it does not wait for certainty
Repeated waves of emerging infectious disease have shown something important in the data: the measure that truly lowers incidence during an outbreak is not the invention of a vaccine, but strict physical isolation. The real strength of a vaccine is to prevent the same infectious disease from spreading again on a large scale; when an outbreak has just begun, it cannot take effect in time.
Statistical inference is, to a large extent, a compromise between absolute rationality and practical decision-making. Precisely because of that compromise, it has produced a vast amount of knowledge. That knowledge is not guaranteed to be logically true in any final sense. A better way to describe science is as a methodology for producing knowledge and nudging it closer to logical truth.
Science seeks truth, yet the methods it actually uses are not fully rigorous in the logical sense. Even so, from states to individuals, people still prefer scientific methods when they want to solve real problems.
Partly that is because science, in secular terms, creates economic value and works. Partly it is because non-scientific systems, even when internally coherent, are far weaker as practical guides.
At an even more basic level, an industrially ambitious society will naturally choose science as its guiding method.
Religion preserved reason; industry chose science
If you look at religious doctrine, you can see that religion was once one of the earliest places where the flame of reason was preserved. Christianity, Catholicism, and Islam all played real roles in maintaining knowledge, even if they also edited that knowledge according to doctrine.
In an agricultural society, where survival depends on the weather and the seasons, religious knowledge may be enough to stabilize social life. But in a modern society, retreat into chanting and metaphysics becomes a kind of self-enclosure. Such a society is easily colonized or overtaken by industrialization. Modern life needs industry, not metaphysical grand talk, and the best source of industrial technique is scientific theory.
Put more bluntly: the choice of science over other systems has less to do with pure rationality than with what best serves survival under a given historical situation.
The data from many developing countries do not show a simple pattern in which education rises first and the economy follows. Quite often it is the reverse: economic growth comes first, and educational development catches up later. In the early stages of development, rapid growth is indeed helped by labor-intensive industries. But it also owes a great deal to developed countries relocating manufacturing. Since manufacturing is often standardized, one can buy a production line and get started without deeply understanding the theory, even if scientific theory helped shape those systems in the first place.
But once basic survival is secured, continued development depends on indigenous technological innovation. And this is exactly the area that the frontiers of science often disdain. In China, many basic disciplines have already reached the world frontier, yet the gap in engineering technology compared with developed countries can still be embarrassing. Elegant and elevated, perhaps—but not something you can live on.
Logic belongs to a few; history is made by everyone
Logic—or more originally, logos—is the pursuit of a very small minority. Knowledge, if we are not speaking only of scientific knowledge, is also pursued by a minority. Feeling, or the ordinary happiness of secular life, is what most people actually pursue.
History is built collectively out of everyone’s perception of the world. But history rarely follows a script, and almost never obeys a single theoretical system. Many people think they have discovered the law of how the world works, when in reality they have only found one angle from which to view it.
For the same reason, many things cannot be discussed by simple logical deduction alone. In principle, any logically self-consistent system can be used to explain everything. But the ability to explain does not mean the explanation is unique, and certainly does not mean it is true.
At that point, the best thing to do is often to look at the data. Data itself already integrates a huge amount of unknown information. That kind of synthetic power is rare for people educated mainly to analyze rather than to absorb wholes. Even so, data can only reveal patterns in what has already happened. New knowledge, in the end, usually requires an outlier with enough impulse to disrupt the whole arrangement.
I know that I do not know. Still, it is worth trying. That may be the calmest possible stance toward the world we live in now.