Picture this: You’re seated across the table from your organic-chemistry tutor. She presents you with a particularly tough problem. Exasperated, you force a thin half-smile. The tutor reads your facial cues, senses your frustration, and offers reassurance.
Now imagine this: Your tutor is a camera-equipped computer capable of reading, analyzing, and reacting to your emotions.
The concept is called affect-aware cyberlearning, and it isn’t entirely new. Sidney D’Mello, an assistant professor of computer science and psychology at the University of Notre Dame, has been testing such automated tutoring “agents” for about a decade with his team at the university’s Emotive Computing Lab. But recent research has allowed them to refine their algorithms, and it has revealed new insights into what teaching strategies induce learning.
One of those insights? Confused students learn better.
In experimental results presented this week at the Cyberlearning Summit, in Madison, Wis., Mr. D’Mello and his colleagues calibrated their computerized tutors to intentionally confuse undergraduate psychology students.
In the experiment, each test subject would engage in a three-way conversation with a cybertutor called “Dr. Williams” and a cyberstudent dubbed “Chris.” In some scenarios, Chris and Dr. Williams would agree. In others, they would contradict one another. Sometimes Chris and the good doctor would even present false information.
Meanwhile, the program looked for facial cues that would indicate confusion. After running various scenarios, the researchers found that confusion actually helped some students understand the material better.
In their paper, they wrote: “The present results are significant because they constitute some of the first experimental evidence on the advantage of inducing confusion during learning. The most obvious implication of this research is that there might be some practical benefits for designing educational interventions that intentionally perplex learners.”
Mr. D’Mello is quick to mention that confusion doesn’t work for all types of learning. It’s best to avoid contradictions when students are tackling simple tasks like rote memorization. But for bigger, more theoretical tasks, confusion can provide a helpful jolt. “You need something dramatic to sort of shake them up and make them pay much more attention,” he says. “That’s what emotions do—they direct attention. They tell you something is wrong with your world.”
Through affect-aware technology, Mr. D’Mello hopes both to better understand how different emotions influence learning and to use those findings to build more-sophisticated cybertutors that can “figure out with reasonable accuracy whether you’re bored or confused, and feed that into the decision making.”
Boredom, in fact, is the next frontier of Mr. D’Mello’s research. His team recently developed a computer interface that “can actually tell, by looking at your eye movements, reasonably reliably, the moment your mind wanders.” The next step is to determine how those mental lapses obstruct learning and what kind of interventions a cybertutor could stage to limit them.
Mr. D’Mello doesn’t think affect-aware technology will replace professors, but he does believe it can augment classroom learning. Strategies like intentional confusion normally don’t work well in large groups, primarily because students bring different levels of background knowledge and, thus, different thresholds for confusion. A fleet of cybertutors, he believes, could free instructors to use such tactics without neglecting some students.
“The dream has always been this personalized, one-on-one tutoring that models the tutoring to your knowledge,” he says.
And for those who prefer their confusion in analog form? We still recommend reading Finnegans Wake.