x Abu Dhabi, UAE Thursday 20 July 2017

Mathematics can predict Iran's nuclear strategy, says scientist

A model based on game theory claims 90% accuracy, and says country will develop weapons-grade fuel but stop there.

As Iran and the international community try to negotiate an agreement on the Islamic republic's nuclear programme, analysts are busy examining speeches and news reports and scouring documents in their efforts to predict the outcome. But according to one New York University political scientist, it can be predicted by simple mathematics.

A numerical computer model developed by Bruce Bueno de Mesquita translates data about a given issue into a number on a scale, which gauges the likelihood of a particular outcome. On the issue of the Iranian nuclear programme, for instance, the model put the predicted equilibrium outcome at 118 out of 200. That translates to: Iran will get to the point where it has developed weapons grade fuel, but will stop there. The point of that, Prof de Mesquita said, was "to prove that they can".

The model may sound quirky, but 90 per cent of the predictions it has made over the past 25 years - from the succession of Ayatollah Ali Khamenei as the supreme leader of Iran to the implementation of the Good Friday Agreement in Northern Ireland - have been correct. That accuracy rate has earned Prof de Mesquita's model renown and propelled it from the pages of academic journals to the desks of Central Intelligence Agency and defence department officials.

That is not to say US intelligence and decision-making is based on a mathematical model, just that Prof de Mesquita's model has been considerably more accurate than the agency's own analysts, according to a CIA study. "Generally, [the CIA] go with whatever result they are more comfortable with, but my results help inform debate to improve the prospect that they become comfortable with something they initially doubted," Prof de Mesquita, who is also a senior fellow at the Hoover Institution at Stanford University, said in an e-mail.

"That is what I see as the main role of modelling in informing policy decisions." Prof de Mesquita's academic focus is game theory, which uses mathematics to show how people negotiate with one another. His model is something of an extension of this. When using the model to make a prediction on a given issue, Prof de Mesquita collates information from news and intelligence reports and the opinions of experts in the given area, who put a numerical value on four variables affecting the issue: who has a stake in the outcome of an issue; what the stakeholders publicly say they want regarding this issue (or, what is their "strategically chosen decision"); how important the issue is to the individuals involved, and how much clout the individuals can bring to bear on the issue.

The model takes the numerical values from each of the variables, makes them into a pattern and charts them, producing a figure on a scale, usually of zero to 200. Essentially the model predicts how individuals will come together on an issue or not. The more individuals involved the more possible alliances there are. With 40 individuals involved, for instance, there are 1,560 possible alliances. Because the model is mathematical, and not based on human speculation, it factors in all the possible alliances that people may not take into account. It also excludes such factors as culture and history, which analysts will often take into account when trying to predict something. The conclusion, according the model, is that individuals everywhere act in their self-interest.

Critics such as John Mearsheimer, a professor of international relations at the University of Chicago, say mathematising politics is a dangerous precedent that threatens to sideline traditional political scientists. However, the accuracy rate of the model is difficult to dispute. Prof de Mesquita has a high level of confidence in the predictions of his model, though it varies from case to case. "It depends on how confident I am in the information the experts have - If the data accurately reflect a general understanding of the current situation, then whatever changes take place are driven by the model's logic, in which I have a fair amount of confidence.

"The track record suggests that with good data, the model is accurate about 90 per cent of the time, so that is my degree of confidence." Aside from his own academic work and his consultancy with the CIA, Prof de Mesquita does freelance forecasting for various clients that request his services, including a number of companies from the Fortune 500 list of top US companies. He is currently working on a prediction for healthcare reform in the United States, and in the pipeline is an update on the Iranian nuclear programme issue and the likelihood of Pakistan's success against the Taliban.

He also just did a prediction for The Wall Street Journal on the possibility of a deal between Vivendi and Comcast regarding Vivendi's 20-per-cent stake in NBC-Universal by December 10. Prof de Mesquita said the deal will go through. Shareholders were possibly taking note.
jspollen@thenational.ae