A new study by researchers at the Islamic University of Bangladesh investigated whether software could correctly identify the emotion of a person typing on a computer, simply by how long they held down the keys and how fast the participants typed. They found that the software was actually pretty accurate: correctly identifying the emotion of the person typing approximately 80% of the time (87% for joy and 81% for anger).


The study involved 25 participants (15 to 40 years old) who wrote paragraphs from Alice In Wonderland in the first part of the study, and then wrote “from the heart” for the second part. The computer used contextual clues such as word choice, in addition to the speed of the typing, to figure out whether the participants were feeling one of 7 emotions (which each participant willingly provided after writing the paragraph during the first part of the study).

The program ran in the background and oriented itself to the user through a combination of learning and algorithms designed to read emotion through typing style. It opens the theoretical possibility of emotionally sensitive worms or trojans, which would wait to gather or send data until the user is more being less careful and more honest (as during periods of anger or joy, which were effectively predicted by the software).

Of course, beyond the sneaky and suspicious uses of background emotion-reading software, we can imagine that our email program might warn us before we send an angry email: stopping us from getting too caught up in the moment. Of course, it may also be used as spyware to allow security forces to more quickly identify those with intense political or social positions.

I think this seems to have more risks than benefits, although I guess it could allow automated customer service to better gauge your reaction and know when it would be useful to pull in a human or to apologize. I suppose, like most technology, it will offer us both benefits and also untold risks. What do you think?