Article from New Scientist , which I have for you translated Текст для перевода: ..
MY HOROSCOPE for this week claims that now is the perfect time for me to move or, at the very least, to get organized. I know it’s pointless, but that won’t stop me from dreaming about some way to predict the future.
Perhaps it does exist after all. One self-proclaimed seer believes he has the answer. Bruce Bueno de Mesquita – a professor of political science at New York University and a senior fellow at the Hoover Institution at Stanford University in California. In his new book, “The Predictor,” he describes a computer model based on game theory that he and others use to predict the future with remarkable accuracy. Over the past 30 years, Bueno de Mesquita has made thousands of predictions about hundreds of issues, ranging from geopolitics to the personal affairs of individuals. He claims that the overall accuracy of his predictions is 90%. But how does he do it?
The predictions of Bueno de Mesquita began in 1979 when he, as a Guggenheim fellow, wrote a book about the conditions that lead to war. He developed a mathematical model that can test the choices people might make and the probabilities of their actions leading either to diplomatic activity or to war. Like any other model, it requires data to be validated.
A good opportunity arose when the U.S. State Department asked for his opinion on the ongoing political crisis in India. The ruling coalition was losing stability, and it was clear that Prime Minister Morarji Desai would be forced to resign, with a new prime minister to be chosen from within the coalition.
Since his doctoral dissertation was focused on politics in India, and knowing that war is an extension of politics, Bueno de Mesquita agreed to help. He compiled a list of everyone who would try to influence the appointment of the next prime minister, outlining their preferences and the influence they wielded. He fed this information into his computer program and asked it to predict how the negotiations would unfold, leaving it to run overnight. His own intuition suggested that the new prime minister would be the deputy prime minister, Jagjivan Ram. Many other experts on Indian politics thought the same.
The next morning, he checked the results the computer had given him and was surprised to find an unexpected prediction about a politician named Chaudhary Charan Singh will be the next prime minister. He also predicted that he would be unable to build an effective coalition and would quickly resign. “Clearly, there are two possibilities: sheer luck or, after all, it meant something.”
When Bueno de Mesquita reported the results to a State Department official, he was taken aback. According to him, there was nothing to suggest Singh would win, and the result seemed, at the very least, strange. “When I told him that I used a computer program I created, he just laughed and asked me not to tell anyone else,” says Bueno de Mesquita. A few weeks later, Singh became Prime Minister. Six months later, his government was dismissed. “The model found the right answer, but I didn’t,” says Bueno de Mesquita. “It’s clear that there were two possibilities: either the model guessed correctly by chance, or I discovered something.”
Strange result
Three decades later, it was already clear that Bueno de Mesquita had discovered something significant. The model was used by Bueno de Mesquita, his students, and clients (including the U.S. government) to make thousands of predictions published in hundreds of peer-reviewed journals. These include predictions about whether North Korean leader Kim Jong Il would give up his country’s nuclear arsenal, how the “land for peace” program would affect the course of the Israeli-Palestinian conflict, and which clients of the risk management group were most likely to commit fraud. According to a study, at the CIA, Bueno de Mesquita’s model provides over 90 percent accurate predictions. (British Journal of Political Science) , No. 26, p. 441). He is currently spending a significant portion of his time working at a consulting firm based in New York.
Where does such accuracy come from? What Bueno de Mesquita does not do is predict random events, such as lottery draws. He also does not claim to be able to forecast stock market trends, the outcomes of general elections, or the consequences of financial crises—events where millions of people have a small influence, but no one is capable of moving the market on their own.
Rather, he is limited to “strategic situations” where a relatively small number of people negotiate over a contentious decision. “I can predict events and decisions that involve negotiations or coercion, intimidation, or cooperation,” he says. This includes domestic politics, foreign policy, conflicts, business decision-making, and social interactions.
His main tool is game theory, which uses mathematical methods to predict people’s actions in situations where the outcome also depends on the decisions of others. “It’s a beautiful explanation for a fairly simple idea that people do what they believe is in their best interest,” says Bueno de Mesquita.
Game theory was invented in the 1940s. John von Neumann и. Oscar Morgenstern It was originally based on games where players tried to anticipate the behavior of other players or their responses, when all participants were honest and inclined to cooperate. In the 1950s, a mathematician… John Nash , the hero of the movie “ Games of the Mind “, he created a more realistic theory in which players can intimidate, lie, bluff, or go back on their word to achieve their desired outcomes. A classic example is the prisoner’s dilemma (see description below). Bueno de Mesquita uses Nash’s assumptions: players act in their own interests and will do everything to achieve what they want — or at least to block an undesirable outcome.
In its simplest form, the model works as follows. First, Bueno de Mesquita decides how to frame the question— for example, whether Iran will develop nuclear weapons. Then, he creates a list of everyone who could influence that decision and assigns each of them a value from 1 to any number, say 100, in each of four categories: what outcome they desire; how important they consider the issue to themselves; how strongly they want to reach a consensus; and what influence they have.
At this stage, the “negotiations” begin. Let’s say we have five players: A, B, C, D, and E. To reach a result, each player’s position is compared pairwise with the others. When A is paired with B, for example, A must decide whether to support or oppose a key proposal (“Iran should develop nuclear weapons”), or whether to suggest a counter-proposal, taking into account B’s position and the likelihood of gaining support from C, D, or E. B either agrees, negotiates, or intimidates in turn, always considering the positions of the other players. Once all possible combinations have been played out, each player sorts through the various proposals or demands they have received and assesses their confidence in any threats against them. Players can thus change their position on the issue at hand. Ultimately, the model allows for the calculation of the group’s overall position as a number from 1 to 100. This option is considered the “outcome.”
If we have only 5 players, we have 120 possible interactions—each player interacting with each of the remaining players in both directions (5 × 4), multiplied by the positions of the remaining three players (3 × 2). However, the complexity of the calculations increases rapidly as the number of players grows. With 10 players, we have 3.6 million potential interactions. A typical scenario that needs to be predicted involves 30 to 40 players, while Bueno de Mesquita has not addressed problems involving more than 200 participants.
In game theory, one of the key factors determining the success of a model is the quality of the input data: you reap what you sow. To achieve high-quality data, Bueno de Mesquita actively consults with experts in the field.
According to the political scientist Nolan McCarty from Princeton University This is the real strength of such an approach. “I suspect that the success of the model is largely due to the fact that Bueno de Mesquita works very well with the input data; he is a highly knowledgeable person and a widely respected political scientist. I am skeptical about the idea that the modeling apparatus provides the kind of intellectual power that he claims.”
Colleague McCarthy’s from Princeton, economist Avinash Dixit He agrees but warns, “Experts can be wrong, as we have recently seen in the case of the financial crisis.”
Dixit points out another issue with the results produced by the model. We are dealing with ambiguity in interpreting the obtained results. For example, if we asked for a prediction on whether Iran will develop nuclear weapons and received a score of 120 on a scale from 0 to 200, what does that mean? Will they make a bomb or not? It’s not so clear. I think that precise answers can be misleading, and game theory theorists should be more modest and openly discuss the uncertainty that is inevitable in such calculations and the results obtained.
Bueno de Mesquita acknowledges that the results require expert interpretation, but he says there is no ambiguity. “The question of the weights is not simply whether a bomb will be built or not,” he says. “Rather, they define intermediate points: the number 120 regarding the Iranian issue means that Iran will move towards obtaining weapons-grade fuel, but will not build a bomb.”
Bueno de Mesquita is currently working on a new and more complex model using Bayesian game theory, which also takes into account players’ beliefs about other players and allows for outcomes in scenarios with imperfect or incomplete information.
“The old model – especially its complex version, which I used in 1979 – was accurate 90 times out of 100,” he says. “The new model outperforms the old one in terms of results and the accuracy of the path leading to those results.” In February, he presented a report at a meeting. Associations of International Studies with a detailed explanation of the performance difference between the two models.
So how good is the new model? Bueno de Mesquita recently used it to make a forecast about the political situation in Pakistan. Working with a group of students, he explored how willing the Pakistani government would be to actively combat Al-Qaeda and Taliban militants on its territory, and how the U.S. government could influence their actions.
Aiming at Terror
In January 2008, students gathered data on all players, including the USA, Pakistan, President Pervez Musharraf, and other leading Pakistani politicians. They speculated that the USA offers foreign aid to persuade Pakistani leaders of the need to combat terrorism, and that Pakistan would seek to maximize the amount of possible aid from the USA.
The model predicts that in order to achieve maximum cooperation with Pakistan, the U.S. will have to contribute at least $1.5 billion in 2009, which is double the projected budget for 2008. In response, Pakistan will pursue terrorists at an 80 out of 100 level, but no more. In other words, the leadership will make significant efforts to reduce the terrorist threat, but will not eliminate it entirely. “The Pakistani government is not foolish,” explains Bueno de Mesquita. “They know that the money tap will be turned off if they destroy Al-Qaeda and the Taliban movement. So they will contain the threat and reduce it, but not eliminate it completely.”
The result? According to Bueno de Mesquita, the U.S. government allocated $1.5 billion in foreign aid to Pakistan in 2009, and the Pakistani leadership is purposefully targeting militants at the expected level. “We have performed very well,” says Bueno de Mesquita.
With such a powerful tool at his disposal, there must be a temptation to use it for personal gain. Bueno de Mesquita admits that he has received several unofficial offers. In 1997, representatives Mobutu Sese Seko Recently, the ousted president of Zaire (now the Democratic Republic of the Congo) was asked to figure out how to maintain power in the country in exchange for 10 percent of Mobutu’s wealth. Bueno de Mesquita warned the U.S. government about this.
However, he used his model to help his friends, as well as to assist the San Francisco Opera when it faced financial difficulties.
So what awaits us in the future? One of the latest forecasts by Bueno de Mesquita was regarding the outcomes of future climate change negotiations up to 2050. Unfortunately, he predicts that while the world will engage in tough negotiations to reduce greenhouse gas emissions beyond the framework of the Kyoto Protocol, in practice, these measures may not be adopted by Brazil, India, and China as their influence grows in relation to the European Union and the United States.
The seer was also incredibly inaccurate. In 1992, he was asked to predict which bills were likely to pass through the U.S. Congress after Bill Clinton was elected. It is well known that Clinton planned to push through a healthcare bill, but all 27 of Bueno de Mesquita’s predictions about what would likely be in that bill and which elements would be accepted by Congress turned out to be wrong.
What went wrong? The problem was with the input data. Bueno de Mesquita assumed that the influential congressman, Daniel Rostenkovski …will be a key figure in implementing the healthcare reform. But as soon as Clinton began pushing his plan, an investigation into corruption was launched against Rostenkowski, and he was forced to resign from his position. Bueno de Mesquita was unhappy at the time and now shrugs: “I was ready to stand by my reputation and publish the results before the events unfolded. I’ve never been so ashamed.”
Game Theory in Action
A classic example of a problem in game theory is the prisoner’s dilemma It was developed in 1950 to illustrate a situation where cooperation is the best overall policy, as opposed to pursuing primary self-interests, which leads to far from ideal outcomes for everyone. Naturally, something like this was to be expected.
You and your accomplice have been arrested and placed in separate cells. Each of you is being offered a deal by the investigator: confess or remain silent. If you confess and your partner remains silent, all charges against you will be dropped, and your partner will be sentenced to 10 years. If you remain silent and your partner confesses, he will go free while you will end up behind bars for 10 years. If both of you confess, you will both be sent to prison for 5 years. If both of you remain silent, there is no strong evidence against you, and the maximum you could face is 6 months in prison each. Neither you nor your partner will know what choice the other has made until the final decision is announced.
The dilemma is that, regardless of your partner’s actions, it seems better for you to confess. Thus, the rational choice is to make a full confession. However, when both of you confess, the outcome is much worse than if you both had remained silent.
Sanjida O’Connell is an editor in New Scientist’s opinion section