The Perfect Teacher is Coming — and It’s a Data-Fueled Robot
An engaging personality. Clear lesson objectives. Possession of the coveted RateMyProfessors.com “hot or not” chili pepper. All are attributes of a good teacher.
But what exactly makes a great teacher — or even, the perfect teacher?
With over a decade as a student — and dozens of professors, lecturers, and teaching assistants to match — I had several case studies to analyze in my attempt to answer this question. Time and again, two features stood out to separate the truly memorable teachers from those not-so-much: comprehensive knowledge in their field or exceptional emotional acumen.
The former, lacking the latter, was typically enough to make me stand in awe of a professor — I revered their unadulterated intelligence, and even tolerated any impatience or curtness simply for the chance to interact with such a gifted mind. The latter, lacking the former, made me feel comfortable in the classroom — my thoughts and ideas were valued, and my concerns and frustrations went noticed.
Unfortunately, very rarely did these two characteristics coincide. But what if we could build a teacher that unfailingly managed both?
With the rise of the educational technology industry, this idea is not implausible. Indeed, few doubt the effectiveness of tutoring systems when it comes to possessing extensive knowledge of any particular subject matter: at least for introductory courses, the curriculum remains relatively static and is easy to program into the software. As evidence, countless tutoring systems flood the market, and are successful at presenting information clearly and accurately, every time. Students can depend on consistently correct and standardized information and avoid the undesirable variance in competence that individual teachers bring.
Perhaps the most common criticism of tutoring systems, however, is that they lack the social aptitude to create meaningful relationships with students. It is true that digital tutors fall flat on this front — they fail to anticipate student confusion and are not dynamic or astute enough to manage an individual student’s emotions.
This could all change, however, with the integration of new technologies into tutoring software. Educational data mining, machine learning, and artificial intelligence is an emerging interdisciplinary field, concerned with analyzing and employing data to better understand and serve students. Although relatively new — the Journal of Educational Data Mining begins its analysis of the state of EDM research with papers from 1995 — the field has progressed significantly in recent years.
But how can data and algorithms — something so inherently rigid and logical — produce a socially- and emotionally-literate teacher? It is common in the field to note the so-called “academic emotions” — confusion, frustration, boredom, and engaged concentration, to name a few — as underlying student motivation. It follows, then, that if we can build an intelligent tutoring system that can recognize and regulate these emotions in its students, then we've built a socially-apt digital teacher.
Researchers are hard at work building this emotionally perceptive architecture. A study at the University of Nottingham’s Learning Sciences Research Institute, for instance, tackles the boredom problem by designing a machine-learned model of students’ off-task behavior. Their model collects easily-accessible data (e.g. time spent between answers, the amount of times a student requests help) and finds correlations between this data and being off-task. Using this information, the system can redirect the student toward more engaging tasks, as a teacher might.
Similarly, researchers at Carnegie Mellon University’s Computer Science Department are studying student help-seeking patterns to curb the confusion and frustration problems. The cognitive tutor used in the study collected data regarding success, failure, and error frequencies and correlated those with a student’s tendency to seek help. By better understanding the conditions under which students use (and abuse) hints, the intelligent tutoring system can teach itself when to offer help and when to withhold it, just as an astute teacher can discern between a student genuinely stuck and a student simply fishing for answers.
Now, while intelligent tutoring systems are becoming more emotionally-apt every day, this is not to say they will replace teachers. In fact, NPR reports that teaching has as little as a 0.4% chance of becoming automated by computers within the next twenty years. The profession requires too much creativity and resourcefulness, something computers are not quite ready for.
Instead, teachers can use these systems to supplement their weaknesses. A teacher with low emotional perceptiveness can use the system to gauge which students get frustrated easily. A teacher with shaky content knowledge can rely on the system to offer clear instruction.
But a teacher without the chili pepper? They'll just have to deal with that on their own.