By Analyzing Data from Over 48,000 Telegram Channels, Artificial Intelligence Had Predicted the Iranian Presidential Elections Outcome
- Vahid Hajjehforoosh
Journalist, reporting on information technology startups
"Nobody knows everything, but everyone knows something."
That is not just an old saying, but the manifest of the Wisdom of Crowds Theory.
Professor Marcus du Sautoy from Oxford University performed an experiment to empirically test that extent to which the aforementioned approach to decision-making finely illustrates reality. To this end, he put 4,510 jelly beans in a jar. In turn, he asked 160 people to visually estimate the number of jelly beans present in the jar.
Of the entire 160 people, only 4 or 5 people guessed a number close to the actual number of the jelly beans; however, not so close as to be seriously considered.
What Professor du Sautoy did next was to add up all the numbers guessed, and divide them by the number of people who speculated. The result came to 4,514, a figure only 4 digits away from the real count! (Nobody guesses right, but everyone get it right).
Professor du Sautoy believes that the submitted guesses were only speculations, but collective decisions can be of greater efficiency than the ones stemming from a small body of like-minded people, in other words: wisdom of the crowd.
On May 14, 2017, a call to submission went viral in thousands of Telegram channels for a contest that would invite people to submit their prediction of the 12th Iran presidential elections' outcome in the country. Within three days, hundreds with various educational backgrounds and levels signed up to participate. Two days later, 330 teams were getting ready to partake in a guessing contest unprecedented in the history of both information technology and politics in Iran.
After the registration deadline, Majid Zarrin, the mastermind behind the artificial intelligence contest, provided the participating teams with the required data for programming and analysis. The data was collected by the contest administrator NetMine Company through crawling over 48,000 Telegram channels, each with over 1,000 members. As presidential campaigns continued throughout the week, new data was frequently collected and given to contestants.
Each team was supposed to write a code for a program based on artificial intelligence algorithms, and announce the outcome of the presidential elections before 11 am on the day, May 19, 2017.
Out of the 330 contending teams, 36 managed to finish programming within the allotted time and present the final result to the administration.
But the administrating team was not done. Within 6 hours, they developed an algorithm based on the Wisdom of the Crowd Theory and set to analyze the findings of other teams, reaching at an ultimate prediction: a result that less than 24 hours later would have missed the real figure by a margin of only 0.7%.
"Around the time of American presidential elections in January, most polls projected a victory for Hillary Clinton," says Zarrin, who hold multiple remarkable titles and scientific achievements in the field.
"However, an artificial intelligence system in Mumbai, India, had correctly predicted that Donald Trump would win Tuesday’s presidential elections in the United States."
Developed by the founder of Genic.ai startup, Sanjiv Rai, in 2004, and unveiled in 2004, the system called 'MogIA' has already gotten it right in four presidential elections.
Zarrin, who claimed first place in the global Angel Hack competition, says Sanjiv Rai has collected and analyzed data from 20 million public platforms including Google, Facebook, and Twitter.
"It is only the first time in Iran that artificial intelligence goes beyond commercial applications and into the society and politics to predict election results," he noted.
The immense reception for this contest signals a bright future for artificial intelligence in the country as well as startups active in the field. Trade, culture, society, and politics can all benefit infinitely from achievements of this particular field which is currently on the highest edge of innovative technology across the world.
Result of the contest to predict the outcome of the recent presidential elections (2017) in Iran, developed based on artificial intelligence algorithms
A comparison between the real outcome of the presidential elections and those predicted by contending teams via artificial intelligence
Educational levels of contestants