What machines teach us about how we flirt
I’ve always been a fan of the MIT Media Lab, and in BLONDES I write about Media Lab director Alex Pentland’s development of a machine that tracks “vocal body language” — pitch, speed, the space between words, and so on. Tested in a speed dating study, the machine had an impressive accuracy rate. About 70 percent of the time it could predict whether a man and woman was interested in each other, and whether they’d say yes to a second date.
Now it turns out that researchers at Stanford are following suit, in what may someday be the next big iPhone app.
Computer scientist Rajesh Ranganath and his colleagues developed a “flirtation-detection machine” that — based on prosodic, dialogue, and lexical clues — can detect a speaker’s intent to flirt with up to 71.5 percent accuracy. (In contrast, men were only 56.2 percent accurate when assessing if their date was flirting with them, and women were only 62.2 percent accurate.) Results were derived by testing the machine’s analysis against the self-reported intentions and perceptions of men and women on a speed date.
Without going into the technical details of the study, here are a few of the findings:
1. [When flirting],men ask more questions, and use more “you” and “we.” They laugh more, and use more sexual, anger, and negative emotional words. Prosodically they speak faster, with higher pitch, but quieter [softer voice] (lower intensity min).
2. [When flirting], women use more “I” and less “we”; men use more “we” and “you”. Men labeled as flirting are softer, but women labeled as flirting are not. Women who say they are flirting have a much expanded pitch range (lower pitch min, higher pitch max), laugh more, use more “I” and “well,” use repair questions (Excuse me?) but not other kinds of questions, use more sexual terms, use far less appreciations (Wow, That’s true, Oh, great) and backchannels (Uh-huh., Yeah., Right., Oh, okay.), and use fewer, longer turns, with more words in general [than do men].
3. Both genders convey intended flirtation by laughing more, speaking faster, and using higher pitch. However, we do find gender differences; men ask more questions when they say they are flirting, women ask fewer, although they do use more repair questions, while men do not.
Interestingly, the device was able to predict flirtation in men with greater accuracy than in women (79.5 percent versus 68.0 percent). In BLONDES I write about evolution and men’s overperception bias; essentially, guys are wired to think women are into them when they are not. I wonder if this is also true of the machines men make.