Imagine you’re a political strategist in 1992. Some consultant walks in the door and says they have access to the aggregate, real-time opinions of millions of Americans. He describes the power of his tool as sifting through the raw, unedited and highly emotional opinions and thoughts from every segment of the American population, and he can analyze it for you. He can predict the future by reading minds. Sounds crazy.
That consultant would have been full of it, obviously. He would have been equally ridiculed in the 2004 presidential election, but what about 2008 or 2012? Social media today presents the possibility of aggregating these broadcasts and creating some sort of prediction. But is it credible?
The uprisings in the Arab world have caused the latest buzz around social media and the power of analyzing its aggregate. Of course, most attention is paid to the use of social media to organize and communicate quickly and many are still asking the useless question of “does it matter”. Internet pessimists, or social media pessimists, should be sitting in dark corners cursing Malcolm Gladwell for making such discussion popular.
Stop asking whether it has an impact. Ask where it’s going to have an impact next. Even if it’s not being used to start a revolution, it represents the opinions and thoughts of millions of the world’s citizens. That’s something worth looking at.
I’m getting to the point of my studies at LSE where I need to write a dissertation. I’ve got pages (in Evernote, of course) filled of notes and ideas on what could possibly be entertaining enough to spend about 5 months obsessing over. Well, this is it.
I used social media analytics in my last job to help clients begin to understand public sentiment about their brand or stance on legislation. We monitored trends over time, sentiment, frequency of discussion, demographics and targeted key-influencers. It’s all rather standard, but surprisingly foreign to the DC crowd.
These tools were useful, but they only told a fraction of the story. After all, “sentiment” was decided by a computer program, not a human, and it only felt confident to code it about 14% of the time. Does this sample provide an accurate subset of sentiment for the remaining, un-tagged data? It’s not clear. Does the sentiment applied to conversations actually reflect the nuances of the author’s intent? Again, unclear.
Artificial intelligence (AI) is improving rapidly, just see IBM’s Watson on Jeopardy. But, this computing power isn’t widely available yet. Software-based social media analysis is a booming business with dozens of reputable companies making huge profits. Firms specializing in more traditional research methodologies using the same data are a bit less common, but still in demand, and far more expensive.
I hope to ask whether social media analysis can predict the outcome of political elections, or at least get a good idea of when it’s wrong. I’ll take some time over the coming months to write about what I’m finding. I don’t reach a very large audience here, but I’d welcome feedback from whomever might visit. It should be a fun process, answering this question. I’m excited to get moving.
Photo: Neoformix (rather standard graph of Twitter mentions of Toyota brand over time with popular keywords called out)