#iranElection: Quantifying Online Activism

Added July 2010

For the final year of undergrad in college, I wanted to narrow down to an insanely specific degree and study the impact (if any) of Twitter on the protests following the Iran Election. I use Sandor Vegh’s categorization scheme to break any piece of particulate conversation (or tweet) into one his three categories of messages in a larger online activist instance (awareness/advocacy, organization/mobilization, action/reaction). By analyzing the data to see the breakdown of messages on a quantitative level, I was attempting to understand the predominance of data in one of these categories, which could in turn give us a rough idea of a data-based solution on how the Iran Election was impacted. While the paper that I wrote eventually came to no major conclusions (as most papers tend to), through the process, I realized that the job of collecting and analyzing the data, while difficult, could be repeated for any other type of study.

In a small way, answering questions like this for a particular platform (in this case, the low-hanging Twitter-fruit), we can start to get at the role of the internet in all our lives (kinda-sorta), which is hugely important and very often lacks hard numbers. For this reason, I split the time working on the thesis and working on a piece of code that automated all the work I had done on my thesis. This way, if anyone wanted to come back and try thing their own way with their own information, there wouldn’t be any issue.

As I’m finished with this project, I am opening up everything for other researchers – namely, the big chunk of data I collected. Links are as follows:

Short version of the paper

Full Paper

Source Text Files of Iran Election data

Full data set

All tweets from the ‘StopAhmadi’ account

Back