Artificial Intelligence Will Serve Teachers, Not Make Them Obsolete

By David Lariviere

One of the biggest hit shows on television these days is HBO’s “Westworld,” a sci-fi drama about a futuristic amusement park hosted by incredibly sophisticated human-like robots.

A major misconception of Artificial Intelligence, commonly known as AI, in general as well as when it’s applied to education in particular, is that it will render human beings obsolete.

Andreas Oranje, a Principal Research Director at ETS, disagrees. “It’s not as omni-powerful as pundits would like people to believe — this scary, uncontrollable thing that would become a life of its own,” says Oranje. “The notion that computers take over the world and try to control all humans — these are great apocalyptical stories for entertainment, but they are kind of ridiculous.”

For all practical purposes, Oranje argues, AI will actually perform many of the mundane tasks in the classroom, such as scheduling, grading multiple-choice tests, and capturing and tracking progress data. Some schools are already implementing these systems.

Far from being threatened by having AI assume these roles, Oranje, who has worked for ETS since 2001, thinks teachers will ultimately embrace and champion new technology. “Nobody is going to reminisce that manually grading multiple-choice questions was so great,” Oranje states.

“The minute we automate, few want to go back,” he explains. “What happens is that we now have new expectations that are more complex and add more value. No teacher will lose their job because of AI. We will come up with new norms and new things we couldn’t even imagine before.”

In fact, AI will be liberating for teachers, Oranje predicts, allowing them to offer their students increasingly personalized learning experiences.

Although there are certainly many ways for students to learn individually today, the set-up of 25–30 children to a teacher in a classroom might not be going away anytime soon, according to Oranje, and personalized learning has a long way to go, especially as an automated system. “Personalized learning sounds really great for sure and I hope it continues to evolve,” he says.

In a recent presentation, Oranje made a bold prediction about where educational testing and AI will be in a decade. He postulated that testing as we know it now will be extinct. “The idea of an artificial event where you stop and test won’t make sense anymore,” he explains. “Assessment should still be deeply embedded inside the activity of learning, but the systems should be smart enough to accurately connect how you interact with current and previous learning materials to what you know, and what you should be working on next. It should be able to adapt to your learning needs at that time, while helping you build confidence to master increasingly more complex skills and gain deeper expertise.”

So where does that leave a leading testing service like ETS? Oranje thinks ETS will continue to play a major, albeit very different, assessment role. The more assessment directly relates to and includes applying the actual skills of interest, the more optimal learning can be informed and accomplished, according to Oranje. “This is already widely used in medical and military training.”

There are many ways to have more direct assessment, especially in digital environments. For example, tools like virtual and augmented reality are especially about getting closer to the real thing, says Oranje. “I mean you could learn from a textbook the makeup of water and describe it on paper, but wouldn’t it be far more interesting if you could pull together two H molecules and one O, and actually see what water looks like at the molecular level? Or in learning about the solar system,” he continues, “if you could actually experience how the earth spins around the sun, to be able to walk around the solar system and move things around — those are the kinds of things in which those tools will actually help deepen understanding,” he states.

Building sound cognitively based evidence models is where developers can play a pivotal role, Oranje thinks, in the form of simulation-based assessment in combination with natural language processing and speech tools that come much closer to the real thing than a multiple-choice test would.

The problem in designing accurate models is trying to encompass the wide range and nuances of performance levels and the complexity of the subjects students learn. “AI will be of help and even necessary for some of this, but it will always be a step behind,” explains Oranje. “What happens is that as efficiencies are gained from the application of AI, immediately new needs will arise because we raise our expectations and standards for instruction and we create new requirements that we were not able to attain or recognize before.”

For AI to make a lasting foothold in schools, however, Oranje thinks that stronger collaborations and partnerships between ed-tech, education research and educators need to occur. “Then we can create useful platforms for schools to get on that are reasonably standardized, interoperable, that talk to each other, and are in the cloud,” he says. “We’ll probably see that in five years, but what I’m mostly worried about is that gaps in accessing the tech will get a lot wider. That’s where we have a role at ETS to bring equity in that we develop all our tools for everyone and we make sure they are accessible to everyone.”