For some education experts, “personalized” learning isn’t only about figuring out how to customize the content of a lesson for each student, but also about how a student is taught. Many students are zoning out even when the instructional content is more or less right. They’re just not in the mood to focus and study.
One researcher in this field is Sidney D’Mello, a computer scientist at the University of Colorado, Boulder, who is studying how students’ minds often wander when they’re learning through software.
Mind wandering is an odd thing to study. It’s an internal state in our brains and we may not always be aware of exactly where our thoughts are roaming. D’Mello’s approach was to install inexpensive eye trackers on the computers in an Indiana high school classroom. Then, as students were learning biology through a software program, he intermittently interrupted them through a computer prompt, inquiring whether they were zoning out. Think of it as an annoying survey that kept popping up on the students’ computer screens.
For each answer, D’Mello and five research colleagues recorded what the students’ eyes were doing: how their eyes were blinking, staring and moving across the computer screen. We’re talking thousands of data points for each student. Then the researchers employed a machine learning algorithm to figure out which eye patterns were typically associated with mind wandering and which weren’t.
The student surveys were riddled with inaccuracies. Many students didn’t answer truthfully when the computer asked if they were zoning out. Others may not have been aware that they were doing it.
To check that their machine-learning algorithm was correctly separating the focused mind from the wandering mind, D’Mello and his colleagues looked at how much the students in his experiment were actually learning through the educational software, which also tracked students’ progress. According to his hypothesis, kids who were zoning out more should be learning less because they’re not focusing on their work. Indeed, D’Mello found that the students whose eyes more often matched the “zoning out” patterns learned less biology than the students who exhibited “not zoning out” patterns. And that’s what gave him and his colleagues confidence that they’d figured out a way to detect mind wandering. The results were published last year by the Association of Computing Machinery. (The study, co-authored with Jim Brockmole at the University of Notre Dame and their students, was financed by the National Science Foundation.)
D’Mello found that mind wandering is difficult to notice just by watching someone. A zoned-out student could be looking straight at the computer screen as if his eyes were gazing straight at the teacher.
“You’re fixating longer when you’re mind-wandering,” D’Mello explained. “You’re responding less to what is happening. It’s almost like a blank stare kind of thing but it happens in milliseconds.”
When students are paying attention to instructional software, their eyes are bouncing around more. Students who are engaged in active learning absorb the material in patches. They fixate on a word or an image, encode it on their brains and then move on. “When you’re zoning out, you’re just fixating,” said D’Mello. “You’re not moving on.”
(The picture of blue dots at the top of this story attempts to give a visual image of the difference. Each dot represents where a reader’s gaze is focused, and the lines show where the reader’s eyes move to next. In the top picture, where a person’s mind is wandering, the blue dots are slightly larger, showing how the reader is staring at words a bit longer.)
After categorizing various eye patterns as “zoning out” or “not zoning out,” the researchers were able track students’ eye movements in real time to predict when their minds were probably wandering. They found that students were zoning out between 20 and 25 percent of the time while they were learning by software. That’s much better than you might expect. D’Mello said that previous research has shown that our minds typically wander 30 percent of the time when we’re reading or as much as 40 percent of the time when we’re listening to a lecture. A separate study of elementary school classrooms found that kids were “off-task” more than a quarter of the time. By comparison, the students in D’Mello’s study were more focused.
However, he cautioned that his experimental results should not be seen as an endorsement of instructional software in general. Students might have been on their best behavior simply because they knew they were being observed. Or students might have been more engaged because the instructional software was new to them — a novelty that could wear off over time. However, it’s also possible that interactive software, such as the Guru Tutor used in this experiment, is relatively more engaging. This intelligent tutoring system is designed to emulate human tutors. An animated character was constantly asking students questions and responding to their answers, in a sort of modified Socratic dialogue interspersed with activities.
Now that researchers believe they can detect mind wandering, what do we do with it? One idea is for instructional software to monitor mind-wandering in real time and combat it in the moment. For example, the animated tutor could say, “Let me repeat that…,” give a student a pop quiz or change the topic.
While the researchers say they have no intention of eradicating daydreams*, solutions like that frighten me. Do we really want to curb mind wandering? It’s associated with creativity, and perhaps a bit of mind wandering is needed to come up with big thoughts. Some mathematicians purposely turn their attention to something irrelevant when they’re stuck on a problem and then the solution magically pops into their heads. Do we really want to create classrooms where cognitive scientists are somehow nudging teenagers out of every daydream back into protein synthesis or verb conjugations? Perhaps the best use of this research would be for evaluating instructional software, pointing out when and where computerized learning is a bore.
But what I find fascinating about this research is how data scientists have come to a conclusion that contradicts human intuition. You often hear teachers say that they don’t need data to tell them what their students know. Well, this research points out that it’s hard for teachers to know when students are really absorbing something just by looking at their faces.
*Clarification: This article was modified from an earlier version to make clear that the researchers aren’t seeking to limit all mind wandering, but only cases when students are exhibiting persistent and problematic inattention that’s affecting their academic performance.