Here’s a pretty interesting site I recently came across: TweetPsych. It uses two linguistic analysis algorithms to pull out patterns in peoples’ Tweets. The algorithms are the Regressive Imagery Dictionary (RID), and Linguistic Inquiry and Word Count (LIWC). RID I’ve never heard of, to my immediate memory, but LIWC we came across in the How to be Happier article, where James Pennebaker used it to analyse journal entries, looking for the best way that disclosive writing could be used as a therapeutic tool.
If TweetPsych, created by Dan Zarella, does what it says on the tin, it’s a great piece of programming. It compares an individual’s Tweets to a baseline set of about 1.5m Tweets, to see where they stand out. Here are my results. It says I talk about cognitive process and positive emotions and stuff, and at the end gives a list of other people who think like me. They were all guys and one was a rocket scientist.
At the moment, TweetPsych is in the early stages, but it has some potential. What I think would be really cool, would be if Dan could validate the data against measures that already exist. For example, get a few hundred Twitter users to complete a well-being questionnaire, like this one, and see how it matches up with TweetPsych’s positive word analysis. At the moment he’s analysing the last 1,000 Tweets, which could cover a long time period. If a short amount of Tweets could correlate well with a validated measure, that could track all kinds of things over time – happiness in relation to world events, for example.
That one’s has been done, actually, but again they haven’t compared the content of social media to established measures, which would be better. There’s no doubt a link between how someone uses language and other things like personality, but how strong is it? And how strong once that language has been diluted into typed words, let along 140 character limit? Maybe there’s something there, maybe there isn’t; one doesn’t know until one tests it (I think that’s the first time I’ve ever used ‘one’).
If he hasn’t done that, it’d be better if it wasn’t touted as “Psychological Profiling”, in my opinion, because that can give people the wrong idea! For example, Tyler Hayes, a media consultant and blogger from Minneapolis, says of TweetPsych: “Ahh… the day has arrived. Psychology has finally wiggled its way into social media,” and describes TweetPscyh as a “psychological profiling & assessment tool.”
But, as Tyler goes on to say, psychology in the official sense hasn’t arrived in social media yet; at least not through Dan Zarella. TweetPsych is absolutely not a psychological profile tool. Technically speaking, it’s about as psychological as a tarot card reading, and it’s the type of thing that gives people the wrong impression of what psychology is. But the technology is great, and could be used for the power of good – if the results were validated properly. Maybe Dan’s really planning to do something scientific with it – he does seem the type. If so, good luck!