Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
SAF Archives  search ask the scientists in the classroom cool science
scientists from previous shows
cool careers in science
ask the scientists

Photo of Prof. Cassell Affective Computing

In this story, Rosalind Picard demonstrates how she's trying to give computers the human-like ability to recognize emotional cues and use them in interactions. Rosalind responded to viewers' questions about the amazing new capabilities of computers by answering questions as part of an Ask the Scientists panel following the premier of the Scientific American Frontiers Special Inventing the Future. Here are viewers' questions and her answers.

Q How does the amygdala act like a fire alarm? What actually happens in your brain when you get scared?

A A great explanation of this process can be found in Joe LeDoux's article: "Emotion, Memory and the Brain," Scientific American, June 1994, pp 50-57.

Q When you get to the point of creating a prototype of a computer that responds to emotional cues, which emotions do you think will be among the first this computer will respond to? I guess another way to ask this question is which do you think will be the easiest emotions for a computer to recognize and which will be the most difficult?

A We don't know for certain yet, but it looks like affective states such as stress or frustration will be easier to detect than states such as interest/curiosity. Some of the hardest emotional states to recognize may be the "cognitive" emotions such as shame or guilt -- especially when what is going on in the person's thoughts is essential to understanding their emotions. Thought recognition is much harder than emotion recognition.

Q Can you predict if this new generation of "responsive" computers will be programmed to respond to the emotions of a specific user, to a general range of human emotions or perhaps something in between - responding to general emotional cues with specific programming possible for an individual user?

A I expect it will be quite individualized, and will perform best with such tailoring. Humans are best at reading emotions when they know you well. I expect your computer will have to "get to know you" and your individual preferences before it gets good at responding to you in a way you really like.

Q It struck me that some of the technology you're using is similar to the polygraph test currently in use. Is that so, or did you begin anew to create these new computer capabilities?

A Yes, some of the measurements we use -- galvanic skin response, respiration, and heart rate -- are also used in polygraphs. Our emphasis is different, however. Polygraph tests are usually given in an adversarial situation -- to try to recognize lies or deception. Our focus is on situations where a willing user and computer are fully aware of each other's abilities and want to communicate openly. In addition, we use a greater variety of sensors coupled with pattern recognition techniques to facilitate this communication.


Scientific American Frontiers
Fall 1990 to Spring 2000
Sponsored by GTE Corporation,
now a part of Verizon Communications Inc.