"People often ask: Why didn't Didi send me the nearest car? This begins with a long-standing economic question..."
Text / Ba Jiuling (WeChat public account: Wu Xiaobo Channel)
What is the difference between global optimality and local optimality?
Some classical economists believed that the pursuit of the maximization of individual interests can add up to the maximization of group interests. But Paul Samuelson points out that this is a "synthetic fallacy" – the mistaken belief that what is true in the local context can be true at the global level.
Samuelson's nephew, Lawrence Summers, a former President of Harvard, gave an example: When an audience member stood up in the room, he saw it more clearly, could it be extended to the whole? When all the spectators stood up, no one could see clearly, and it was even more uncomfortable.
In the field of operations research, the "greedy algorithm" hopes to achieve the optimal outcome by selecting the current optimal at each step. However, in doing so, it is inevitable that there will be problems of quick success and short-term profits, greed for small and big losses.
A snowball, facing a very steep and short track and a very slow and long track, how to choose? Greedy algorithms choose those that grow fast initially and then come to a standstill early. And Buffett will say: Life is like a snowball, the important thing is to find very wet snow and long slopes.
People gradually found that global optimality is not so easy, and even summarized some classic error models:
Tragedy of the Commons
What is the outcome of a piece of public grassland that everyone has the right to use, but no one has the obligation to maintain? Every shepherd would bring a large herd of sheep to graze (locally optimal) – they knew that the meadow could not withstand it, and they were even less willing to fall behind – and finally the commons ceased to exist (the worst of the whole), which was the "tragedy of the commons".
The tragedy of the commons has many derivative cases, such as the misuse of antibiotics:
"Antibiotic effectiveness" is a public resource that has benefited all of humanity since the discovery of penicillin in 1928, but no one is responsible for this effectiveness.
Therefore, in order to treat the disease once, many people use antibiotics in large doses. However, the more people who do this, the more drug-resistant superbugs will be produced, and the antibiotics will gradually become ineffective.
Every year, China produces and consumes nearly half of the world's antibiotics, and while most of them are used to raise livestock, that doesn't change the fact that misuse gives rise to superbugs.
Another example is the abuse of adjectives:
"The validity of adjectives" is also a public resource that everyone uses and no one manages. Therefore, in order to express the effect of a good time, many people use the highest level of description at will. However, the more people who say this, the more the meaning of the word will weaken, and the original adjective will gradually depreciate.
Years ago, the word "beauty" expressed praise for a woman's appearance. But the more people who shout "beauty" every time, the weaker the potency of the word. Nowadays, the word "beauty" is almost an everyday title for any woman, and praising beauty requires "big beauty", "super big beauty", "goddess" and so on.
Some would say that these words are useless. The Japanese invented another way to add time or space qualifiers. First the "once-in-a-decade Yicai" Matsui Zhu Rina, then the "once-in-a-millennium beautiful girl" Hashimoto Kana, the Chinese Ju Jingyi was named "the idol of the 4000-year encounter", Oguri Yuki was "the beautiful girl who met once in 20000", and Haruka Toyokawa was "a beautiful girl who met once in 3 million light years"... Space-time continues to depreciate.
Pulling the scale for a long time makes it easier to see the weakening of the meaning of the word. In the "Left Biography" more than 2,000 years ago, "hunger" means that it is more serious than hunger, and it is impossible to eat independently, and "starvation" means that it is really starving. Today, more than two thousand years later, at four o'clock in the afternoon, you can hear "I'm hungry, I'm starving!" ”
In addition, from the global climate, marine resources, biodiversity, to the space on the sidewalk, there are problems with the tragedy of the commons.
Free-riding problem
The tragedy of the commons changes slightly, and it becomes a free-rider problem: when a public good can be used without paying, everyone tends not to pay (local optimal), but if everyone does not pay, there is no public good available (the worst overall situation).
There are also many derivative cases of the free-riding problem, such as public epidemic prevention:
If there is only one person in the country who has not been vaccinated, then he is almost impossible to get sick like a vaccine, but if there is a certain proportion of such people, the epidemic will break out.
Different infectious diseases have different free-riding behaviors to "collapse the border". Seth Berkeley, head of Gavi, the Vaccine Alliance, has said: "Once measles vaccine coverage is below 95%, there will inevitably be a large outbreak of infection. ”
In 2000, the United States declared the elimination of measles. Between 2010 and 2017, more than 2.5 million children in the United States were unvaccinated against measles, the highest in a high-income country. As a result, in the first half of 2019, the measles epidemic made a comeback in the United States, and the number of infected people hit a new high in nearly 25 years.
U.S. Department of Health officials said the outbreak could have been avoided entirely if vaccination rates had not fallen.
Another example is the online loan credit report:
A group of online lending companies, in order to improve the level of risk control and reduce the cost of risk control, decided to share users' credit and loan records. However, everyone thinks that I don't have to do credit reporting seriously, just use the data of other people's homes on the platform... The result is to pit each other's teammates and concentrate on detonating thunder.
In addition, from UN affairs to team work, there are debates about free-riding.
Prisoner's Dilemma
A and B conspired to commit the crime and were caught by the police, but the police did not have enough evidence to accuse, so they interrogated the two separately. At this time, the fate of A and B faces four possibilities:
If neither of them confesses, each is sentenced to 1 year; A confesses and denounces the other, and B is silent, A is released and sentenced to 10 years; if A is silent, B is sentenced to 10 years and released; if A and B confess, each sentence is 5 years.
It is not difficult to see that "whether the other party confesses or not, he will confess" is the most favorable choice for each prisoner (partial optimal), and both sides will jointly fall into the end of betrayal (the worst of the whole world) - this is the "prisoner's dilemma".
The Prisoner's Dilemma also has many derivative scenarios, such as corporate price wars:
If A and B do not fight a price war, they will maintain the status quo; if A fights B does not fight, A will grab more markets; if A does not fight B, B will grab more markets; A and B fight, not only will the market share remain unchanged, but the revenue pressure on both sides will increase significantly.
Another example is the national tariff war:
Neither country A nor B fights a tariff war and opens its doors to do business; if A fights B does not fight, the competitiveness of enterprises in country A is enhanced; if A does not fight B, the competitiveness of enterprises in country B is enhanced; if A and B fight, this is what we are experiencing, the trade relations between China and the United States are damaged, and the economies of various countries are also affected. (Of course, we had to fight)
Interpersonal relations, intercity development, international politics, biological evolution... Cases of prisoners' dilemmas can be found everywhere.
Global optimal?
So, what does global optimality look like? Not all scenarios can achieve global optimality, because global optimality often requires "God's perspective" and "top-level planning".
To the larger extent, the climate change agreement, the tariff and trade agreement, the central bank's credit reporting system, the national epidemic prevention plan, and the high-speed rail shipping plan are the pursuit of global optimality; to a small extent, Didi's vehicle allocation plan is the pursuit of global optimality.
Many people have questioned why there is a car nearby, but Didi has to send a distant single? - For the sake of global optimality.
Didi is essentially a coordinator and matchmaker, using big data algorithms and economic laws to try to make more passengers hit the car during peak periods, so that drivers can have orders during the trough period.
Big data algorithms mean that getting more people to get to the car faster is not necessarily an individual user hitting the nearest car, but the overall waiting time for everyone is shortened.
The laws of economics mean that through time-sharing pricing and other methods, balance and match the supply and demand sides to achieve market equilibrium. During peak periods, bad weather or hot spots, more drivers are encouraged to get out of the car, improve the response rate, and also encourage users to choose ride-sharing, preferential access, chauffeur-drive and other travel modes to alleviate the lack of capacity.
The Internet, big data, and artificial intelligence make global optimization possible. Didi is such a practitioner, striving for better travel for more users.
The author of this article | Cats have medicine men | When the value is edited | He Mengfei
Responsible Editor | He Mengfei | Editor-in-Chief | Zheng Yuanmei | Illustration | Nanke a guest