Emotional Intelligence Developed For Computerized Tutors

Beverly Woolf, a research professor of computer science, says the introduction of the emotion sensors helps the scientists respond to how people actually learn. “Emotion and cognitive functions are strongly correlated,” Woolf says. “So if you improve the social intelligence of the computer, students respond the way they would to another person. Sensors allow the computer to identify students who pay attention and those too tired or bored to learn. Using these cues, the computer provides individualized instruction. ”

Woolf is part of a team of scientists that includes Ivon M. Arroyo, research scientist, and Andrew Barto, professor of computer science at UMass Amherst, and Winslow Burleson from Arizona State University. The tutors they have designed teach geometry and algebra to high school students, but can be adapted to any subject, she says. The work is funded by grants from the National Science Foundation.

Woolf says the non-invasive sensors replicate what top-notch human teachers do in the classroom to engage their students. “Master teachers devote as much time working on a student’s motivation as they do on straight teaching,” she says. “They understand that students who feel anxious or depressed don’t assimilate information properly.”

The sensors they are developing include a camera that views facial expressions. Woolf says certain looks on a student’s face or how they tilt or hold their head are strong indicators of their level of interest in what they are doing. There is also a posture-sensing device in the seat of a chair to measure movement. This measures the amount of fidgeting, or stillness, other indicators of interest and concentration on the task.

There is also a pressure-sensitive computer mouse that can tell how hard the user is pushing down. Previous research has shown that users who find an online task frustrating often apply significantly more pressure to the mouse than those who do not find the same task frustrating, Woolf says.

In addition, a wireless skin conductance wristband worn by the student shows how activated the person is. A certain amount of arousal is a motivator toward learning and tends to accompany significant, new, or attention-getting events, she says.

The combination of the emotion sensors and mathematical subject matter are presented to the students during a period of up to two hours, Woolf says. During each session, the computer analyzes the information it gets from the sensors and adjusts how it presents the information. Sometimes, that means halting the program and offering the student an alternative activity to reignite interest. Or it could involve having the computer go back and revisit material that the student has failed to master.

Sensors also pick up when students try to “game the program” by randomly choosing answers or hurrying through the problems. When that behavior is detected, the computer tutor responds, in a friendly manner, and asks them to slow down or read more carefully.