Feel festive, cheerful and blessed around the holidays--but then slide into the doldrums in the first weeks of the New Year? Financially illiterate and then suddenly started blogging about how the ups and downs of the stock market impacted you emotionally? Felt patriotic--or depressed--when Obama was elected?
Oh yes, the internet knows.
Talk about following the zeitgeist: Computer programmers Sep Kamvar and Jonathan Harris have spent more than four years collecting some 12 million emotions posted on Internet blogs. Turns out we're a pretty predicable bunch: Patterns of the calendar, news events and even the weather influence how we say we feel. And as an increasing number of bloggers worldwide share their lives publicly, we're developing a new relationship with computers, our fellow bloggers and ourselves.
And this holiday season, you can track your emotions in their strikingly beautiful, glossy gift book, We Feel Fine (Scribner, Dec. 1), that uses sophisticated computer science to underpin its findings about modern human emotion. The brainchild of Kamvar, a professor of computational mathematics at Stanford University, and Harris, a systems designer, the data collected comes from a program that scans all blogs every few minutes and extracts the sentences that contain "I feel" or "I am feeling." Since blogs often have public profiles, the duo was able to determine the gender, age, and location of the people expressing these emotions to boot.
Kamvar said he and Harris hope to tell both macro stories of emotional trends, including informational graphics and maps, and micro stories of individuals complete with a photo and corresponding feeling. "We wanted the reader to be able to seamlessly transition between the high-level statistics of emotion and the individual stories that make up these statistics."
"Our [original] intent was to show that there was beauty and humanity in the web," said Kamvar of their 2005 website born as social networking media was coming of age. " As time went on and we collected large amounts of data, we realized that we were also building an archive of emotional history."
As We Feel Fine: An Almanac of Human Emotionhit the shelves, Kamvar, a college classmate, sat down with me for a Q&A:
What are the top 10 most common feelings 2006-2009?
• Better (as in "I feel better" or "I am feeling better now")
• Bad
• Good
• Guilty
• Sorry
• Sick
• Well
• Comfortable
• Great
• Happy
What about September through November of last year when the market was tanking? Did the feelings reported differ from the average?
We saw a rise in anxiety during the fall of 2008. Also, prior to the crash, there were very few people in the general public who felt anything at all about the economy - the word "economy" was rarely used in sentences that had the words "I feel" in them. In the fall of 2008, that all changed.
How about on the day after Obama was elected?
On Election Day and the day after Election Day, the feeling "proud" spiked to the highest levels that we'd seen in the 4 years that we've been collecting emotions. The number of people who expressed feeling "excited" and "patriotic" also spiked. There was also a smaller spike in people feeling "depressed."
We enjoyed analyzing the types of emotions elicted by the candidates over the course of the campaign. By doing that, we were able to compute real-time approval ratings.
Any other big event that showed a dip or spike in particular emotions?
Michael Jackson's death had a pronounced effect on emotions in the U.S.-- which was one of the biggest dips in happiness that we've seen. [And] notably, MJ's approval rating also rose dramatically after his death.
Are men's reported feelings different from women's reported feelings?
Absolutely. Women express their emotions far more frequently than men, and have a more nuanced vocabulary to describe them. The feelings expressed by men tend to be more individualistic (e.g. "I feel proud") while the feelings expressed by women are more interconnected (e.g. "I feel loved").
You analyze how people feel during certain weather conditions--why?
Because we could. We knew the time and location when each feeling was expressed, so our program could look at existing weather databases and figure out the weather in that time/location.
The findings are not surprising - sunny weather is correlated with more happiness than gloomy weather, but other factors (like age) are much stronger influencers on emotion.
People in different age brackets report different levels of frequencies of feelings (older people are more likely to feel gratitude than younger people, for example). Do you think that is true over time, or are we seeing particular generational and cohort differences here?
Some of the biggest emotional differences come from people in different age groups - for example, teens' feelings involve angst and adrenaline more often than people in their 50s. Given the nature of the differences, I think the reasons are deeper than specific generational dissimilarities.
As an example, we noticed that the meaning of happiness shifts as people get older. Younger people tend to associate happiness with excitement, while older people are more likely to associate happiness with calm. We followed this up with some research with Cassie Mogilner at UPenn and Jennifer Aaker at Stanford, where we found that this was due in part to a greater sense of presence as people get older. By influencing young people to think about the present, we could make them define happiness as older people do, and by influencing old people to think about the future, we could make them define happiness as young people do.
What is the most surprising finding of the book?
Americans express their feelings about sex less than any other English-speaking country. And Americans express their feelings about God more than any other English-speaking country.
What are some of the methodological weaknesses of using blog data to chart emotions?
Well, there is definitely a population bias - bloggers tend to be younger and more tech-savvy than the overall population. You can account for the age bias by sampling equally from different age groups, but it's harder to account for other cultural biases.
Despite that, it's surprisingly accurate, in large part because you can perform studies with 2 million people, while most other methods take a much smaller sample.
In general, I like to use this method in conjunction with other methods to get a fuller picture.
How would you like to see this research used in the social sciences?
I think studies of online human expression (via blogs, Twitter, social networks, etc.) are a useful complement to existing methods in the social sciences, particularly because of their large scale and low cost. Ideally, I'd like research like this to play a similar role to the microarray in biology, which made gene expression experiments cheaper and quicker.
If you could have someone take away only three things from the book, what would you want them to understand/feel or learn?
• Our own biggest takeaway is that people are more the same than they are different. The book draws from people around the world, in all age groups, and the emotions are surprisingly universal. We came away from this book with a feeling of empathy and self-reflection, and we hope that it has the same effect on others.
• We'd also like this book to provoke some thought about the relationship between humans and computers. Millions of people are sharing their emotions with their computers on a daily basis, and slowly, we are teaching computers what it's like to be human.
• And finally, this book was possible because of a cultural shift, where people are increasingly living their lives in public on the web. This shift has both wondrous potentialities and its own dangers. But it is increasingly our reality and it's useful to reflect on both.