Rosalind W. Picard welcomed both skeptics ("I was one myself a few years ago") and enthusiasts to a lecture that celebrated the publication of her book, Affective Computing (MIT Press) and challenged her audience to reconsider the role emotions play in our own lives and the role they might play in the "lives" of computers.
Dr. Picard, the NEC Development Professor of Computers and Communications and associate professor of media technology at the Media Laboratory, opened with a quick survey of research on emotions and their role in human decision-making. Studies reveal that too much emotion wreaks havoc on reasoning, but, surprisingly, too little emotion also take its toll, she said. Of computer scientists, she quipped, "We're not known as empathically savvy people. Yet empathy -- reading and adapting behavior to affective cues -- is critical."
Empathy is especially important, she noted, in teaching and learning. Three emotions -- interest, distress, and pleasure -- form a natural cycle associated with learning, and if a mentor can adapt his or her behavior when recognizing them, more effective interaction results.
Now that distance learning has become more commonplace, Professor Picard asked, can computers become mentors in this way? Further down the road, can computers become like affectively-clued-in house pets, figuratively purring or trotting in with the evening paper and a pair of cyber-slippers?
Professor Picard acknowledged that she was once a skeptic, but no longer. Some computers now have the ability to express and recognize affect or emotion, and she said her research at the Media Lab indicates these abilities could become more commonplace.
Computers have been made to recognize emotional states as revealed by facial expression, vocal intonation, posture and gestures. Computer-generated speech has also been modulated affectively. "It's easier to come up with negative examples of emotions," said Professor Picard. "It's really hard to make them sound joyful right now."
Less visible indicators of emotion such as pupillary dilation, respiratory changes, heart rate, pulse and temperature can all be recognized through skin conductivity. In fact, some research shows an 80 percent accuracy rate for identifying three different emotions, she said.
"Still images are hard to look at and recognize. Computers like exaggerated expressions, with the mouth going up or down, as in anger, disgust or surprise," she said. "Computers like someone speaking to them as if they're speaking to a foreigner."
Professor Picard then provided her criteria for judging whether an affectively equipped machine is actually a conscious one. She noted that the intuitive "feeling of knowing," the ability to regulate emotions, and the ability to utilize emotions as people do when planning activities to capitalize on good experiences, are all aspects of consciousness that still elude researchers.
Concerns about affective computers -- including invasion of privacy, anthropomorphizing computers (leading, perhaps, to a computer-rights activism analogous to animal-rights activism) and the possibility that computers may learn to feign affect and "manipulate or deceive us" -- were also raised.
Professor Picard closed her lecture with a summary of graduate students' works in progress, including wearable computers such as the Blood-Volume Pulse Earring, a galvanic skin response unit inside a shoe, and an affective camera pendant that gathers video continuously but saves only those images associated with states of high arousal.
Video footage of Confusion-Sensing Glasses which detect eyebrow-furrowing (intended for use in distance learning) offered both a view of the future and a note of Saturday Night Live-style humor. Questions following Professor Picard's talk sought more information on applications of her research and explored the implications of generalizing about human emotion.
Professor Picard is delivering another lecture, "Toward Machines That Can Deny Their Maker," this afternoon at 4:30pm in Rm 34-101. Hers is the final lecture in the fall series, "God and Computers: Minds, Machines, and Metaphysics," coordinated by Dr. Anne Foerst of the Artificial Intelligence Laboratory.
A version of this article appeared in MIT Tech Talk on December 10, 1997.