Machines do a lot of things well, but one thing they can’t do is understand how people feel—or can they? Rosalind Picard, and other researchers at the Massachusetts Institute of Technology (MIT) are working to change that through novel uses of sensors.
Picard, who gave a keynote address yesterday at the Design East conference in Boston, is the director and cofounder of the affective computing research group at the MIT Media Laboratory, a group dedicated to bridging the gap between human emotions and technology. Her team is experimenting with ways to use electrodermal sensors, which measure the electrical conductivity of skin, to gauge sympathetic arousal, otherwise known as the body’s fight-or-flight response.
When the body sweats, the conductivity of the skin increases. Sensors can measure that conductivity, and the data they capture can provide insight into stressful experiences. For example, data Picard captured from MIT students who tested the sensors showed that electrodermal activity was low during resting activities, such as watching television. It spiked during stressful activities, such as studying and taking exams.
Picard said the technology can help provide better understanding of people who can’t otherwise express how they are feeling outwardly, such as children with autism who have a limited ability to speak. She showed video of a girl with autism who was wearing electrodermal sensors, along with a graph of the corresponding electrodermal activity data they captured. As the child appeared to become agitated, the graph spiked; when the child played on a swing, an activity thought to induce calm, the line leveled out.
The team’s research has also led to some unexpected discoveries. For example, they predicted that electrodermal activity would fall during sleep, but instead data showed a spike in electrodermal activity levels during slumber.
Picard said the most surprising discovery came when one of her students took the sensors home over Christmas break to test them on his brother who has autism. The data he captured showed an abnormally high peak in electrodermal activity that Picard initially thought was due to a hardware malfunction. She asked the student to check his notes to see what happened during that time, and as it turned out, the spike occurred just before his brother had a seizure.
Further research has shown that wrist-worn electrodermal activity sensors, combined with accelerometry biosensors, can accurately predict convulsive seizures 94% of the time. The data is highly correlated to EEG, but the sensors are a lot easier to wear, Picard said.
Future applications for the sensors could include detecting use of illicit substances by people being treated for addiction and helping to predict the onset of disorders such as Rett syndrome, a nervous system disorder that leads to developmental reversal that mainly affects girls. Picard and Rana el Kaliouby, a research scientist at the MIT Media Lab, have cofounded a company, Affectiva, to commercialize the technology in a wrist-worn product called the Q-Sensor, which features electrodermal activity sensors, a 3-axis accelerometer, and wireless connectivity.
Picard’s affective computing research group is also studying ways to use video to measure physiological parameters and gauge emotions. One application to come out of the group is Cardiocam, a technology that can determine a person’s heart rate through color changes in their face captured via web cam. Its accuracy compares with that of a traditional finger-worn heart rate monitor. The technology is currently available as an iPhone app, Cardiio, and the team has also placed it behind a half-silver mirror for home-use applications.
—Jamie Hartford is the associate editor of MD+DI. Follow her on Twitter @readMED.