Desktop computers to counsel users to make better decisions

What if a personal computer knew how its user is feeling?

That computer on your desk is just your helper. But soon it may become a very close friend. Now it sends your e-mails, links you to the Web, does your computations, and pays your bills. Soon it could warn you when you’re talking too much at a meeting, if scientists at Sandia National Laboratories’ Advanced Concepts Group have their way. Or it could alert others in your group to be attentive when you have something important to say.

Aided by tiny sensors and transmitters called a PAL (Personal Assistance Link) your machine (with your permission) will become an anthroscope – an investigator of your up-to-the-moment vital signs, says Sandia project manager Peter Merkle. It will monitor your perspiration and heartbeat, read your facial expressions and head motions, analyze your voice tones, and correlate these to keep you informed with a running account of how you are feeling – something you may be ignoring – instead of waiting passively for your factual questions. It also will transmit this information to others in your group so that everyone can work together more effectively.

“We’re observing humans by using a lot of bandwidth across a broad spectrum of human activity,” says Merkle, who uses a Tom Clancy-based computer game played jointly by four to six participants to develop a baseline understanding of human response under stress.

“If someone’s really excited during the game and that’s correlated with poor performance, the machine might tell him to slow down via a pop-up message,” says Merkle. “On the other hand, it might tell the team leader, ’Take Bill out of loop, we don’t want him monitoring the space shuttle today. He’s had too much coffee and too little sleep. Sally, though, is giving off the right signals to do a great job.’”

The idea of the devices has occasioned some merry feedback, as from a corporate executive who emailed, “Where do we get the version that tells people they are boring in meetings? Please hurry and send that system to us. A truck full or two should cover us.”

More seriously, preliminary results on five people interacting in 12 sessions beginning Aug. 18 indicate that personal sensor readings caused lower arousal states, improved teamwork and better leadership in longer collaborations. A lowered arousal state – the amount of energy put into being aware – is preferable in dealing competently with continuing threat.

The focus behind the $200,000 effort, funded by Sandia’s Laboratory-Directed Research and Development program, is to map the characteristics that correlate to “personal-best” performances.

“The question is, how do we correlate what we observe with optimum performance, so that we improve your ability and the ability of your team leader to make decisions? He can’t tell, for example, that your pulse is racing. We’re extending his ability,” says Merkle.

Those concerned about privacy – who see this as an incursion similar to HAL’s, the supercomputer that took over the spaceship in the movie 2001 – can always opt out, he says, just like people choose not to respond to emails or decline to attend meetings.

But in a sense, he says, the procedure is no different from that followed by people who have heart problems: they routinely wear a monitor home to keep informed of their vital signs.

“In our game, what we learn from your vital signs can help you in the same way,” he says. “It’s almost absurd on its face to think you can’t correlate physiological behavior with the day’s competence.”

After gaining generic maps of individual performance, the information would be linked in a working group through a program called Mentor.

No theory yet exists to explain why or how optimal group performances will be achieved through more extensive computer linkages. But Merkle doesn’t think he needs one.

“Some people think you have to start with a theory. Darwin didn’t go with a theory. He went where his subjects were and started taking notes. Same here,” he says. Merkle presented a paper on his group’s work at the NASA Human Performance conference Oct. 28-29 in Houston. “Before we knew that deep-ocean hydrothermal vents existed, we had complex theories about what governed the chemistry of the oceans. They were wrong.”

Now it’s state-of-the-art to use EEG systems to link up brain events to social interactions, he says. “Let’s get the data and find out what’s real.”

The tools for such a project – accelerometers to measure motion, face-recognition software, EMGs to measure muscle activity, EKGs to measure heart beat, blood volume pulse oximetry to measure oxygen saturation, a Pneumotrace(tm) respiration monitor to measure breathing depth and rapidity – are all off-the-shelf items.

“We give off so much information. But our only current way of interacting with a computer is very limited: through, essentially, a keyboard and mouse. So the limitation of my computer’s ability to help me – this increasingly complex, wonderful machine with its ability to recognize intricate patterns – is its inability to recognize complex patterns in me.”

Is all this really necessary? He answers with some humor, “Not at all. You can always ride a horse; you don’t need an automatic transmission.”

Asked whether this mechanistic view of human behavior can be accurate when many athletes, scientists, and artists have described themselves as feeling poorly yet made unusual gains in their work, and polygraphers have been unable to locate spymasters based on similar reading of vital signs, Merkle replies: “I would not say that we have a mechanistic view, unless one considers studying precedent to be a mechanism. Based on a history of prior performance, we make a prediction on likelihood of suitability for current tasks. It’s no different from making decisions based on baseball statistics: against left-handed batters in the last 200 night games, this person hits .207, so pinch hit the .298 person for him.”

Further work is anticipated in joint projects between Sandia and the University of New Mexico, and also with Caltech.

“In 2004 we intend to integrate simultaneous four-person 128-channel EEG recording,” says Merkle, “correlating brain events, physiologic dynamics, and social phenomena to develop assistive methods to improve group and individual performance.”

To complement this applied research, Sandia is supporting a $50,000 graduate fellowship to study the neurology of learning processes under the Caltech Campus Executive program.

The Sandia project teamed with small business to produce the apparatus. Dave Warner, Steve Birch, and Tim Murphy of MindTel LLC, of Syracuse, N.Y., delivered the prototype with off-the-shelf components and custom software, based on an inexpensive networked PC platform, under budget in only 71 days, says Merkle.