What is cause and effect?

Let me start with my undergraduate thesis. In the discussion class last semester of senior year, our topics were statistics and causality, and graph model learning. Generally speaking, traditional methods can only get the correlation between two variables, but can't get the causal relationship. On the other hand, graph model is a frontier method to obtain causality by using statistical data. Maybe these sentences are too theoretical. Let's give a classic example to illustrate.

Smoking can cause lung cancer-this may be a truth in most people's minds, but in fact, many statisticians have always denied this conclusion. This has been debated in the medical and statistical circles for half a century, and there seems to be no difference so far.

The statistician's reason is simple. There is indeed a strong correlation between smoking and lung cancer, but this in itself is not enough to show that smoking will "cause" lung cancer, because it is impossible to infer who caused whom from two sets of highly correlated data by traditional statistical methods. Statisticians put forward a simple model, assuming that there is a gene that "causes" smoking and lung cancer at the same time, that is to say, people with this gene are easy to be addicted to cigarettes and suffer from lung cancer at the same time, which leads to a high degree of overlap between these two groups of people, which also leads to a high correlation between smoking and lung cancer, while smoking itself has no causal relationship with lung cancer.

We observe, analyze and explain in order to predict and control. However, if this gene exists, even if a person with this gene is not allowed to smoke or asked to quit smoking, he is still prone to lung cancer; For people without this gene, even if they smoke more, they are unlikely to get lung cancer. If this conclusion holds, many people's lifestyles will change. The experimental and clinical data are not obviously on the doctor's side, and even some large-scale clinical data show that the probability of lung cancer in the smoking cessation group is even greater than that in the non-smoking group.

Of course, you might think that anatomy will tell you what actually happens to smokers. Their lungs are black, not smoke? But in fact, lung blackening is a "phenomenon" of cancer. People who smoke have this pathological phenomenon in their lungs, and the existence of this gene cannot be denied. Because it is also possible that genes cause lung diseases, which develop synchronously with gene-induced smoking behavior, thus showing this consistency. Another possibility is that genes lead to smoking behavior, and genes themselves lead to lung cancer together with smoking behavior.

I know maybe you still have a lot to say, and you are even angry at my statement. How can smoking not cause lung cancer? It's sheer nonsense! Wait a minute, let me say two things. First, I have already said what logic is. Our conclusion-the first reaction comes from emotions and beliefs, and then logic begins to serve beliefs, so be calm and don't be kidnapped by your own inertial thinking and emotions. Secondly, let me give a simpler example to illustrate the relationship between relevance and causality.

It is said-probably a joke-that when a czar was in power, there was a plague all over the country and many people died. When he heard someone report to him that the province with the most deaths was the province with the most doctors, he immediately ordered the execution of all doctors in the country! Starting from the positive correlation between the number of doctors and the number of deaths, the czar simply and rudely concluded that doctors caused deaths, which is ridiculous, isn't it? There are at least three simplest possibilities.

First of all, the situation inferred by the tsar does exist in theory. The more doctors, the worse, but the following two may be more reasonable.

The third situation is also more reasonable than the first one, that is, due to geographical and communication reasons, some provinces have serious plague disasters, many diseases and many deaths, and some provinces have light disasters, few diseases and few deaths; The disaster attracted a large number of doctors to participate in the rescue. If so, then more deaths in turn lead to more doctors.

That is to say, two variables A and B (the number of doctors and deaths, or smoking and lung cancer, etc. ) shows a correlation, which may be that A leads to B, or B leads to A, or that A factor C (population and gene) leads to A and B at the same time. If we arbitrarily say that smoking causes lung cancer, it is the same as the tsar's way of thinking. Another possibility is that A and B lead to C at the same time, thus showing a positive correlation. This example is relatively simple. Suppose A is the age set of all students in Class One, Grade One, and B is the age set of all students in Class Two, Grade One; These two sets of data are obviously positive, but who do you think is the cause and who is the result? Neither, because they are all about 6 years old, so they are all in grade one. The last possibility is that, in fact, the two sets of data are completely irrelevant (as long as they are not discussed in the context of universal connection), but they show correlation, which is completely coincidental.

In retrospect, cause and effect is not so simple, is it? In the last case mentioned above, we often add causality to completely irrelevant things, which is what we often do in real life. The book Devil's Economics gives an interesting example.

In 1980s, new york was a hell on earth. Juvenile delinquency is rampant, and the government cannot solve this problem by all means. In the 1990s, juvenile delinquency seemed to disappear overnight. Some people think it is because of the patrol and gun control system, some people think it is the effect of education, some people think it is the effect of economic growth, and some government officials have prospered. But the author of "Devil Economics" found that it was actually a judgment twenty years ago that decided this major turning point! In this judgment, the federal court declared abortion legal, so many future criminals-single-parent families, drug addicts, criminals' children-were not born at all! Of course, neither the author nor I think that birth is everything, but statistics show that children with these family backgrounds are more likely to become criminals. This example is just to show that the systems and measures that have been taken for granted for many years have nothing to do with the result of the decline in crime rate.

So why on earth do we like to make mistakes in causality? First of all, because in many disciplines, mathematics and physics do not have a strictly defined causal relationship, and more are statistical concepts. We are also more tolerant to allow mutual causality, virtuous circle, vicious circle and so on; At the same time, considering different objects and putting them at different time levels, we will get different conclusions (I partially covered this issue in the article "Uncovering the Fraud of Corporate Social Responsibility"). Secondly, as I mentioned in What is Logic, the conclusion comes from faith, and logic serves faith. When the conclusion is confirmed, our belief tells us that smoking will definitely lead to lung cancer, so the causal relationship must be the same. In addition, as I said in the last article, the ass determines the head, and everyone is pursuing advantages and avoiding disadvantages. The reason for this or that causal relationship is to prove their correctness. Isn't that why those officials in new york got promoted?

However, what I really want to pay attention to in this article is another level of reasons. That is, imposing cause and effect comes from human instinct. -because it is very easy for human beings to establish the connection and causal relationship between things in their brains, regardless of whether these things are actually related or not and to what extent. This is probably related to the operating mechanism of our brain neural network, which has left a deep mark in every era when human beings constantly know and explore the world.

I like watching sports news on Sina. I often see headlines like "Yao Ming doubles, Rockets win" and "Kaka is absent, Milan is out". It's one fact that Yao Ming has two doubles, and it's another fact that the Rockets won. There is a certain connection between these two events, but there is no "definite causal relationship", because we know that many times Yao Ming won two pairs and the Rockets lost. These two things may not even be positively related, because when Yao Ming is in a stable state, he may always get two pairs, and whether the Rockets can win or not may depend more on the cooperation of the whole team.

But seeing these two sentences, many people's minds will "translate" that the Rockets won because Yao Ming got two pairs, or Yao Ming played a decisive role on the court. When we open the webpage, we find that Yao Ming actually got two pairs in the first three quarters, but he was injured, and finally won with Artest's key shot, we will find that our "translation" is actually unreliable.

The most exaggerated time seems to be the career review of an international. One year he joined a certain Championship team, and the next year the team was promoted to the Premier League. When I read this sentence together, my first reaction was that he must have played well that season, but the following sentence surprised me-players were nailed to the bench all season because of injuries and state reasons. I think many people will feel the same way-or more people don't even realize it.

From the psychological mechanism, this unconscious "translation" comes from the way our nervous system works-because it is not professional enough, I can only say it roughly. Our daily experience will establish a connection between several neurons, and similar repeated experiences will strengthen this connection and make us form inertial thinking. Once we encounter something similar, current will flow through the related neurons, thus reflecting a whole set of thinking patterns in our minds. Many times these thinking patterns are just specious causality.

From the perspective of understanding the world, this thinking mode of forced "translation" is actually an abstract ability. In the long years of human evolution, this ability is a favorable factor to a great extent, because there are always some interrelated natural phenomena in nature-although they are not always in a strict sense of causality. Mastering these laws can really help people adapt to nature better. Therefore, this is the result of natural selection, which has been passed down from generation to generation through culture and education, making many of us believe-naive from now on-that things always happen for a reason.

Children who like to quickly establish connections between different things will be considered smarter by adults. Isn't it? A child is asked to close his eyes. When he opens the gift, he will probably close his eyes the next time he sees the guest. Aren't we all praising him after seeing such a thing? Adults, why not? Many people associate the Wenchuan earthquake with the construction of the Three Gorges Reservoir for no reason, just because of the time sequence. After my two leaders left, someone joked that I could lead. Just one after another, just two samples, people will forcibly sum up the law, and this instinct of desperately looking for cause and effect is really hard to see!

If this is just a joke, but in an era when our ancestors stubbornly believed that dipping leaves in water and sprinkling them on the ground can pray for rain, collective indulgence can promote the high yield of the earth (these two are simulated witchcraft), and tying the enemy's hair to a puppet and burning it can hurt the enemy (this is sympathetic witchcraft), this belief is widespread among human beings. When people find that this method doesn't always work, they fantasize about some moody gods or immortals-witchcraft gradually turns into religion-and focus on pleasing them, attributing the uncertainty of the effect after asking for rain to the mood of God. Because religion recognizes uncertainty, human beings have made progress. When science and technology gradually replace religion and become the mainstream, bringing us more and more certainty and more and more "truth", we are about to put aside uncertainty and skepticism.

But you should know that what is known is always limited, and what is unknown is always infinite. It is very doubtful whether the causal relationship we get through the finite world can be applied to the infinite unknown world. Hayek once said that abstraction is not so much a unique ability of human beings as that the world is too complicated, and human beings have to abstract in order to turn infinitely complicated practical problems into something that can be solved. Because abstraction has assumptions, definitions, omissions and even distortions. The ritual of asking for rain and the law of gravity are both ways to know the world, but we say that witchcraft is "wrong" and Newtonian mechanics is "right". But when we study free fall, don't we also ignore the volume, buoyancy and air resistance? What's more, when we expand our vision to the speed of light, Newtonian mechanics is "wrong" and we can only use relativity. In fact, I think that in some cases, the theory of relativity may be "wrong" and cannot explain the problem. Blindly believing in Newtonian mechanics and everything we think is "universally applicable" may be as foolish as believing in the causality of witchcraft in some new situations if it deviates from our original assumptions.

In the final analysis, there is no absolute objective cause and effect in this world. Like logic, it is also the product of our subjectivity-the so-called Tao can be Tao, which is extraordinary. Different degrees of understanding lead to different causal relationships between discovery and cognition in each era-but the mechanism of seeking rain does exist in the same dimension as Newtonian mechanics, that is, our subjective understanding level. As I said before, we observe, analyze, explain and infer cause and effect in order to predict, control, better adapt to and change the world. The mainstream or popular view may explain some phenomena to some extent, but it may just be popular-just like the example of the decline of crime rate in new york. The spirit of doubting authority, the ability to think independently and the attitude of not arbitrarily imposing cause and effect on things are one of the most precious qualities of human beings in this era. This is my understanding of cause and effect.