Emotion-sensing PCs could feel your stress
- — 22 April, 2005 07:46
Computers that can read and respond to human emotions can be more effective and reliable than computers that do not, according to Rosalind Picard, professor and founder of the Affective Computing Research Group at the Massachusetts Institute of Technology (MIT).
For instance, a health-care organization providing medical advice with interactive applications giving the impression of being empathetic runs less risk of being sued than such an organization using traditional software, Picard predicted in a presentation at the MIT Information Technology Conference Tuesday.
"When people interact with other people who ignore their emotion, who don't empathize with them, who don't look concerned when they are upset; it drives them crazy," she said. "And yet technology is full of that. There's a time to lower your voice, there's a time to be quiet and there's a time to cheer people up. That goes for both people and computers. You always have emotion, if you don't; you're actually very impaired when it comes to interacting successfully with people in the work place and at home."
That conclusion was applied and tested on three interactive computer systems used in different studies. The reports, one of them not yet finished, show or indicate that users prefer software applications that respond with empathy. The study's respondents were asked which of the systems they preferred or their impressions were measured by other means.
In one recent study, using software that interrupted the users with questions or tasks about their health while the users were doing other things, the users were more responsive to the questions asked by the empathetic system. They also found the empathetic system's interruptions less annoying than disturbances from the traditional "cold" system. The empathetic system interrupted users 496 times, but the users estimated 336 interruptions. The non-empathetic system interrupted users 295 times and they estimated the number of interruptions to be 292.
"Showing a little concern makes a big difference," Picard concluded.
Picard's research group depends on funding and she seems convinced that affective computing can prove lucrative.
"When you address people's frustration well, they become more loyal customers," she said referring to an unpublished study by an unnamed company. Although the demand for affective computing is big in education and health care, Picard predicts the automotive industry will be the first to apply truly interactive affective computing to products.
"It will come fast in automobile navigation for safety reasons," she said. "Sensors can decide the driver's emotional condition. A stressed driver might need to be spoken to in a subdued voice or not interrupted at all."
Microsoft, IBM and Hewlett-Packard laboratories have recruited students from the Affective Computing Research Group, which also indicates the companies that are interested in the field.
Picard pointed out that affective computing also can be misused.
"So many people in the automobile industry came up to me and said 'If you can tell when people are stressed, then you can play happy music!' But, no, I don't want to play happy music and I don't want it to say 'Cheer up!' As we've seen, that is stressing and has a detrimental effect on driving," she said.
Instead, Picard said; "The most effective way we've seen to help people manage their frustration is to first acknowledge it and second empathize with it. After you do that, maybe the person is willing to go have fun."
Picard also objected to some use of responding to emotion in current software. She used "Clippit" as an example, the smiling Microsoft Office assistant that performs a little good-bye ceremony when the user clicks on its exit button: "If it winks and dances when you just want it to go away, most people get irritated. You just want to shoot it. But if it instead looked a little more upset and subdued, then I think people might want to give it another chance."
Another affective computing incentive is high insurance costs for health-care organizations and companies, according to Picard. "As we have fewer and fewer medical people to care for more and more people you're going to find more informatic systems interacting with people. The classical engineering approach to clinical medicine is cold. But, now, we make an argument that it's got to give you the feeling that it is caring and empathetic, otherwise you [the patient] are more likely to sue the organization that provides it."
A study on non-verbal behavior by Harvard University professor Robert Rosenthal supports Picard's argument. The study shows that from voice clips of less than 30 seconds of doctors' interactions with one set of patients, it's possible to predict which doctors are more likely to be sued by a different set of patients. The doctors who give the impression of being nice and emotionally responding are less likely to be sued.
Picard often refers to the interaction between dogs and humans in her examples. She also uses that illustration when other researchers point out that no theory of emotion has been accepted across science disciplines. Some argue that one first has to come up with an accepted theory of emotion -- what emotions really are, how they can be detected and defined -- in order to apply affective computing.
"We don't need that," said Picard. "The dog is an example of a system that I don't think has a good idea of emotion, but it is still useful -- not as intelligent as a human being but still able to handle emotions. I think we will develop a theory of emotion by trying to build it." She also thinks intelligent computers are a possible spin-off from affective computing.
Adjusting the software's emotional responses in a long-term interaction is Picard's biggest challenge.
"Empathy works well in the beginning, the first day. It works a little differently the second day and a little differently the third day. How do you put the aspects together with all the other things in life and read it? Affect is an important missing piece but it's not the only piece, we have to recognize it jointly with all the other things it interacts with." She continued: "And how do you respond when they are using the phone in different places? When they are at home, driving their car, crossing a street, sitting in a taxi?"
Picard worries that the industry will take emotionally responsive computer development to an extreme and make everything emotional, happy and smiley.
"Sometimes you need people to get angry and upset, for example at bad software, in order to invent better software -- anger is good for progress," she said.