Emotion detector: What blogs tell us about mood

In the Journal of Happiness Studies: An Interdisciplinary Forum on Subjective Wellbeing (yes, for real, here) a paper by two University of Vermont statisticians claims that the nation's mood can be determined, in part, by the tone of it's blog postings.

Here's a chart from the paper (available here) showing an increasingly happy nation over the last four years (hearts are Valentine's Day and green trees are Christmas):

Of note is the overall upward trend (didn't political acrimony seem high back in 06?), the periodic declines in July/August (summer doldrums/stock angst anyone?), and the incredible sadness apparently felt among the blogogentsia over Michael Jackson's passing.

You can validate the statistical and research methods and conclusions for yourself (here). Essentially, they boil down to an automated index of blogs using a standard list of rank ordered words shown to convey particular emotions. Then, a magic algorithm is applied that enables age and geographic filters, among other things, to be applied to results and, voila`, the Emotion of The Crowds.

So What?

In some regards, the idea of a collective happiness index is similar to what pollsters have been attempting for decades...except that the sample size in the blog survey is much larger and is near realtime (it's automated afterall). It's also less representative of the diversity of the population.

And as fascinating as the idea of a global barometer of happiness might be, the idea highlights the challenge inherent in defining something as uniquely personal as happiness for such a large group. The paper's authors certainly concede that opportunities for improving the approach exist. They even go so far as to suggest that the approach could be more useful as a means of evaluating 'social contagion' and 'predictive theories' of social interaction...which sounds a bit like what stock markets seem to have become.

For marketers, however, it is a useful reminder that high-level trends can be valuable in scenario planning, ideation and in direction setting. It's also useful in reminding us that what may appear to be true and meaningful at the level of the group can break down completely when applied to the individual...that's as true of messaging, positioning, and creative as it is of our own happiness.

For those of you made happy as RUSH completists, here's the Emotion Detector demo tape:

Google's In-Flu-ence: The illness of crowds

In yet another fascinating example of Google's power (both the power of the data they have access to and their power to crunch it), they've announced the availability of Flu Trends.

The premise is simple: track where people are searching on terms associated with influenza (especially symptoms and treatments) to see if there are trends that can forecast influenza outbreaks.

Given the realtime nature of the data, combined with Google as a nearly ubiquitous starting point for information seekers online, one wonders what other wisdom might be mined from Google's data on what the crowd wants to know.

Fire ant sightings? Kudzu migration? Lindsay Lohan sightings? Of course Google trends has been on the scene for years...But predictive epidemiology data? That's where you have to have a baseline to compare the hypothesis against the experiment...and CDC data is the baseline.  Alas, Google searches do appear to be accurate predictors of flu outbreak (see chart).  

The global network becomes the hivemind...I, for one, welcome our new Google overlords' influence.