Even if you’ve never heard of the Bechdel Test, you know what it is. How often, when watching a film, do the female characters have conversations with other female characters about topics besides men? That’s it. That’s the test, portrayed by cartoonist Alison Bechdel in 1985. Since then, legions of pop culture writers have held films and TV shows up to the Bechdel lens (and you can find many of them here). The test quickly became a vehicle not just to discuss individual films, but to analyze how women are portrayed on screen, and what kind of influence that bears on society.
Thirty years later, researchers have built another kind of Bechdel test, one even more fitting for the times. Computer scientists, who will be presenting their research at the AAAI Social Media and Weblog conference this year, designed an algorithm with the same set of criteria Alison Bechdel used for films to analyze conversations on social media. Twitter, it appears, has a Bechdel problem.
“My main goal was to try to test if the representation of gender that we had in movies is represented in society or not,” explains David Garcia, a researcher at the engineering university ETH Zurich. “We found that it had a huge bias, and it was basically an artifact of Twitter.”
In 2013, the researchers started analyzing movie trailers on YouTube and using algorithms to crawl scripts for Bechdel scores. But they didn’t stop there. Garcia and his colleagues then looked at who shared those trailers on Twitter.
In the Twitterverse, the researchers found that women were 30% more likely to share movies that passed the Bechdel test. (This appeared to be proactive, Garcia noted.) But the computer scientists wanted to take their analysis one step further: They wanted analyze dialogue among Twitter users themselves, like one enormous film scene. So, the researchers started combing the thousands of conversations from roughly 170,000 Twitter users who had shared the films. Were female Twitter users talking among themselves about something besides a man?
The bias held strong. Researchers found that only 10% of Twitter dialogues were between women and didn’t mention men, while twice the proportion of male-male conversations remained lady-free. On Myspace, there was actually no significant difference between men and women in this regard–20% of both genders had conversations independent of the other.
So what does this tell us about social media? Some of the bias, Garcia explains, simply has to do with the fact that there are more dudes on Twitter. But Garcia also wonders if Twitter’s public-facing qualities also had something to do with it. Myspace conversations, he notes, are private–maybe on Twitter, women just aren’t as involved in bigger discussions.
A finding like this calls for more research. But either way, it can’t hurt to look at your own “following” list to see how the gender ratio bears out. Tech writer and entrepreneur Anil Dash tried only retweeting women for a year, and it didn’t change much–although he did feel more thoughtful and exposed to realities that he wouldn’t have encountered otherwise. That sort of seems like the point of social media, no?