Using Twitter To Predict Elections—Or The Flu
Thursday, November 15, 2012
We're sorry. This audio clip is no longer available. A transcript for audioclip 15894 has been made available.
Ming-Hsaing Tsou is sure Twitter can be used to accurately predict who will win a presidential election—just not quite yet.
Tsou and Dipak Gupta, both San Diego State University professors, are part of a project called Mapping Ideas. The idea is to track and analyze publicly-accessible websites, as well as social media like Twitter and Facebook, for key words and phrases and then track their spread.
The project tracked the frequency of mentions of "Obama" and "Romney" on Twitter the week before the election. Remarkably, its results closely resemble the polls taken during that last week.
This kind of mapping, which uses various technologies to gather and analyze data, can also track the spread hate groups or terrorism and of diseases or illness, like the flu or whooping cough.
"Ideas spread through the world, and they always take definite paths, from fashion to political extremism," Gupta said. "This project is designed to capture why and how they spread, and what does it mean for the rest of us."
Tsou said the project "aims to test the hypothesis that the spread of ideas is not random, that there are places that are more prone to host these sites (and accept and spread an idea) than others."
For example, when Florida pastor Terry Jones said he planned to burn a copy of the Koran, "immediately it spread like wildfire," Gupta said.
By tracking who was most interested in the news, they found a "hot spot" in Topeka, Kansas.
"We wanted to know why, and we found out that there was another small church that were saying they were going to do the same thing," he said.
Claire Trageser contributed to this report.
To view PDF documents, Download Acrobat Reader.