In the future, PCs, like presidents, may be able to feel your pain. Or at least they'll recognize it, thanks to new software that lets computers observe and react to users' emotions.

Such "emotional intelligence" is the subject of Rosalind Picard's book Affective Computing (MIT Press, 1997). Picard, NEC Development Professor of Computers and Communications at MIT's Media Lab, argues that our work lives would be more productive and humane if computers could read and respond to our moods. Say a new computer-based training program has an employee on the verge of tears. A PC reading his or her expression might respond by slowing the program's pace.

Affective computing software can "sense" emotions in several ways. It might, for example, work with a digital video camera that records when a user smiles or frowns. Picard is also experimenting with glasses that track eyebrow movements and a mouse that senses how hard it is being clicked.

Those devices feed information to a system that contains rules about nonverbal cues (for example, that a furrowed brow means confusion). Since people have different expressions, the software adapts itself over time to a user's signals. Taking into account the task being performed, the computer can then determine the best way to help that user. For example, it could learn to delay the delivery of E-mail from a user's ex-husband when she starts gritting her teeth, indicating that she's in a foul mood. Affective computers will also be able to tailor how they present information on the fly, offering to switch to graphs or charts when users confronted by long lines of numbers start tearing their hair out.

Associate editor Emily Esterson spoke recently with Picard about the potential of empathetic machines:

Q: Why would we want computers to be emotionally intelligent?

A: Most people use computers like tools. But in the next 20 years, more people will delegate more complicated tasks to computers. For example, they may ask software to go out on the Web and find the best deal on a specific piece of furniture. When you're dealing with software that doesn't necessarily wait for you to tell it what to do but that takes initiative and offers suggestions, it behooves that software to have emotional intelligence -- decision-making and prioritizing skills.

Q: How is affective computing different from artificial intelligence?

A: Artificial-intelligence agents don't recognize what a person is feeling. They may watch what you're typing, but they don't know how to apply emotion when it comes to acting on it. In affective computing, the software has the ability to see not only what you typed but also how you typed it, and to know what that means. For example, you might have pounded the keys, which would tell the computer that you're angry.

Q: How will the way we interact with emotionally intelligent computers differ from the way we interact with the insensitive variety?

A: The key thing is that the software has the potential to adapt to a person. As the computer learns about you, it will seem a little smarter each day. It will make work a more engaging and productive experience.