Do We Tweet More When We're in Certain Moods?

More than anything, our current map is a proof of concept. We wanted to explore the power of the Twitter data and decided that the mood idea would be a neat test case. When we saw the results, we were impressed with two things. Firstly, there was a strong daily pattern in the data (the variation is much greater than the error estimates), and second, we saw a significant regional variation. These results were so striking that we decided to share them.

Now, I have a quantitative background, so rather than speculate, my strategy is usually to go out and measure the answer to any question. Since we know the Twitter data contains many biases (for example, tweeters are not a representative sample of the population), what we're working on now is anchoring information in the Twitter data to results obtained from other sources, such as polls and Centers for Disease Control and Prevention data. This way, we can learn about biases and limitations.

Could it be that Twitter users only tweet in a specific frame of mind—a frame of mind that is not representative of their overall state? This is a specific type of bias. One way I might test this hypothesis would be to look at the mood of many other texts (e.g., blogs, Wikipedia, classic literature) and see how tweets stack up compared with these reference points. If the average use of mood words is similar to usage in these texts, I might conclude that the language of tweeters is representative of their other written expressions. Individuals could be evaluated in a similar manner (assuming that they had a statistically significant number of tweets).

If I were to speculate, I'd say that the answer to this question is likely to be related to the individual user of Twitter and his or her personality; some people probably exclude negative tweets (since your boss might not like you complaining about incompetent management, for example), while other people might see Twitter as a place to vent and get rid of some of their excess anger.

Sune Lehmann is a postdoctoral research fellow at the Institute for Quantitative Social Science at Harvard University and the College of Computer and Information Science at Northeastern University.