Computer Program Knows When You're Struggling With Math By Looking At Your Face

Researchers are working to create math programs that watch kids' faces and adjust problem difficulty based on how users feel.

A facial mood-detecting program by Fraunhofer

A recent example of mood-detecting software that's not the same as the face-reading software used in the study of kids doing mathphoto by Jurvetson (flickr)

If only Number Munchers had known when you were getting frustrated.

A new pilot computer program reveals how hard a kid finds math problems just by watching her face as she works, researchers report. In the future, a math game with built-in face-reading technology could adjust itself to individual kids' ability, the program's creators, a team of communications researchers from Tilburg University in the Netherlands, write in a new paper.

Google Scholar searches show that computer scientists have worked to make software that detects people's moods based on facial cues since at least 2002. I saw an MIT project at last year's SIGGRAPH conference, which captured live video of every conference attendee in the room and layered a virtual red, yellow or green cartoon face over everyone's body, depending on his mood. (It was as weird as it sounds.) This kind of science is still in its early stages, however, so few researchers have been able to link it to a really practical application.

This new study gets closer to a real-life use for auto-detecting moods. Some existing computer programs already react to users' abilities. The software that runs grad-school admissions tests, such as the GRE, adjusts problem difficulty based on whether the test-taker got the previous problem right (Thus driving test-takers crazy… Is this question easy because I'm good? Or because I'm really bad?). A face-reading program, on the other hand, would help by identifying emotions that have little to do with math ability, but still have a big effect on how kids learn. Sure, a teacher can identify students' emotions, too. But a teacher can't be there all the time, watching every single child's face. In the Tilburg study, the face-reading software guessed difficulty levels correctly 71 percent of the time. The software isn't a responsive Number Crunchers--but someday it could be, the researchers say.

Hat tip to EdSurge for highlighting the new study. The study appears in the May issue of the journal Computers in Human Behavior. Be sure to visit the paper's link to see adorable photos of kids trying to concentrate.