 |
 |
 |
 |
| |
|||||
|
|
|
Home > News & Events > MITRE Publications > The MITRE Digest > | |||||||||||||||||||
Communication Breakdown: Applying User Models to Improve Team Decision Making March 2006
You dawdle outside the glass-walled meeting room. Your meeting is scheduled to start in five minutes, but the current occupants of the room don't appear to be tying things up. Harry, Assistant Director of Packaging and Shipping, is hunched back in his chair with his arms crossed and his chin swinging back and forth with the stubborn regularity of a pendulum. Greta, Division Manager of Southeastern Sales and Marketing, is leaning across the conference table and stabbing her finger at Harry. Her voice is mounting in volume and pitch, although her words are too muffled by the glass to make them out. Marla, VP of Product Placement, is spreading her hands out before her, palms up, her head tilted to the side. She is speaking softly and slowly to the other two. "This meeting isn't finishing anytime soon," you mutter resignedly. Even through glass walls, people can be plainly understood. As human beings we do not rely solely on words to communicate. We also read meaning into body posture, gestures, and tone of voice. Divining those cues provides us insight into the mood and motivations of those with whom we are speaking, insight that is often critical to successful team decision making. However, it is sometimes a luxury to communicate face to face with team members when solving a problem. In this day of multinational corporations, outside consultants, and telecommuters, meetings are often conducted with participants speaking over the phone or from a grainy video screen. Verbal and non-verbal clues can be lost or muddied, leading to crossed wires, frayed nerves, and mangled agendas. Keeping the Team on Track Brad Goodman, a principal artificial intelligence (AI) engineer at MITRE, thinks he has a solution to putting the "team" back in "team building." Picture an on-line collaborative environment equipped with an AI "facilitator." The facilitator monitors the speech and action of the meeting participants. When it notices that one member is not fully participating, the facilitator gently prods the member—using a text chat window, synthesized voice, or even a simple indicator display—to contribute while reminding the other members to invite the quiet member's comments. When the facilitator senses a participant growing angry or one growing sulky, it again provides feedback that helps guide the meeting back to a more even-tempered atmosphere. "The idea of providing an artificial intelligent agent as a facilitator," says Goodman, "is to try to get the group to work better together, to get the other members involved in trying to resolve that argument or just stepping back and asking questions to get people answering things, to get them moving forward again instead of arguing." Goodman and his own team—Jill Drury, Robert Gaimari, Laura Chung Kurland, and Guido Zarrella—are designing such an artificial intelligent agent as part of MITRE's internally sponsored research program. However, the agent won't be limited to monitoring the behavior of team members during decision-making activities. It could also be used to build a "user model" of each participant based on his or her personality type or communication habits. "The idea is that as people interact more and more, information about them gets revealed," says Goodman. "You can learn about their style of interaction and use that information to promote better small-group dynamics." These models could be consulted during a meeting to interpret the behavior of each participant and guide the rest of the team in responding to the participant. The models could also be consulted when building a team. "You might be able to tell that this person will work best in a team with people of this type," he says. "Maybe this person likes to speak up a lot and dominate the meeting. He or she might not be the best person to put with somebody who sits quietly at meetings and so on." Time-sensitive Team Research The first step in designing such an agent is to teach it to recognize human behavior in a team-building environment. To gather the necessary data for this programming, Goodman is joining forces with another MITRE group investigating team decision-making for time-sensitive targeting. The shared experiment will revolve around a 90-minute exercise in which three people take on the role of a military team engaging in a time-sensitive targeting mission. One member will act as the team leader/intelligence analyst. The second member will act as the sensor expert. The third member will act as the weapons expert. The team members will interact through a computer targeting system. Different teams will be studied under different conditions with every exercise audio- and video-taped. Goodman and his team will then comb the recordings of the experiments for every scrap of information using a set of data analysis tools they developed to navigate multiple streams of synchronized data. From the video Goodman will identify nonverbal cues; the audio recordings will be subjected to voice stress analysis; and the logs of the computer targeting system will be reviewed to determine the effectiveness of each team's collaboration. Communications, both verbal and nonverbal, that led to effective teamwork will be categorized, as will communications that resulted in ineffective teamwork. The intelligent agent can then be programmed to recognize communications in a team setting that should be encouraged and those that should be discouraged. The Future in Collaboration Once the intelligent agent is programmed, the second step will be to incorporate the agent into an automated meeting facilitator. When a team gathers, remotely or in the same room, it can employ the facilitator to maximize its effectiveness. Goodman cautions that the facilitator need not necessarily be as conversational as HAL from 2001: A Space Odyssey. "It could be as simple as the set of meters we designed for an earlier experiment that indicated the interaction style the team was using. For example, when the team agreed a lot, the meter would indicate it, as it would when they argued a lot. By providing team members immediate feedback on their discussions as they move through the decision-making process, such meters might lead a team to reach consensus more quickly." Goodman believes his findings will aid MITRE sponsors in numerous ways, from building teams perfectly suited to their assigned problems to constructing meeting environments that maximize collaboration and communication. "What we're trying to do is to better understand team decision making," says Goodman. "And from there help teams make better decisions." —by Chris Lockheardt Related Information Articles and News |
||||||||||||||||||||
| Page last updated: March 28, 2006 | Top of page |
Solutions That Make a Difference.® |
|
|
