VR is for learning, AR is for doing

Jeff Dieffenbach


The promise of augmented and virtual reality (AR and VR, respectively) is nearly unlimited. AR and VR won’t just be for entertainment, however … they’ll likely revolutionize learning.

There’s a tendency to think of AR (overlaying graphics onto a view of the real world) as a precursor to VR (a completely virtual world). For AR, think Google Glass and Pokemon Go. VR, on the other hand, conjures up more sophisticated, fully immersive video games (Fortnite, anyone) and books/movies such as Ready Player One.

When it comes to learning, I see the line between AR and VR differently. At risk of oversimplifying, VR lends itself to “learning before doing.” That is, one might use a VR simulation to practice a task in a safe space, for instance.

AR, on the other hand, promises to shine in the manner of “learning while doing.” Imagine an aircraft engine technician with a maintenance task in front of her. Rather than toggle her attention between a manual or video and the engine itself, she would be better served having an overlay of the engine schematic and the service steps projected onto the engine. In this way, her actions would be guided by the AR “assistant” without the need to disengage from then re-engage with the work at hand.

VR is for learning, AR is for doing.

Reaching Students One by One

Reaching Math Students One by One” by Tina Rosenberg of the New York Times (Fri Mar 13, 2015) does a great job of telling a story of instructional technology use in a K-12 setting. Such technology offers the promise of better outcomes and lower cost. In the case that Ms. Rosenberg examines, the jury’s still out on both the outcomes and cost questions, but it’s nonetheless worth taking a deeper look.

“For any subject, any room, it can’t be true that one teacher teaching 30 kids is the best way,” says Joel Rose, an education expert who in 2009 worked for the New York City Department of Education. There, he and Chris Rush, a consultant, developed School of One, a method for math teachers to personalize instruction, and brought it into three middle schools.

The school knocked down three walls to create a giant classroom, and Ailene Altman Mitchell, the principal, said she also spent $140,000 to buy a laptop for each child, along with screens and tables.

I wonder if the decision to buy rather than lease was driven by grant funding or another one-time pot of money. The article isn’t clear, but it appears that this might be for 120 students. Spending more than $1k per year per student on technology seems high to me. Laptops or Chromebooks can be leased for less than $250 a year, and infrastructure, instructional software, and professional development shouldn’t add more than $250-$500 per year beyond that.

The three [students] were discussing the errors, and endorsed the method. “It engages kids,” said Tianna. “You don’t get bored listening to the teachers. And the computers make it more fun.” Romel and Danielle said they liked working in groups.

I’d much rather have a student engaged in an average curriculum than disengaged from a great one. That said, I see no reason why we can’t have both.

The first step, now widespread, was the digital lesson, usually a computer game or video. The best-known web provider, Khan Academy, is free. But there are many other sources of digital lessons, from pricey packages created by education companies to rudimentary videos that teachers make at home.

A caution–the article seems to be leaning at least slightly against “pricey packages created by education companies” (but, perhaps that’s just me detecting a bias that isn’t there). Reading skill development software tends to cost on the order of $30-$40 per student per year. Given that schools are spending more than $10,000 per student per year (in many cases, much more than this), it’s hardly fair to call $30-$40 “pricey.”

Here’s a second consideration–in my opinion, care must be taken in using curricula created by teachers. This is in no way meant to slight teachers, only to make the point that any teacher creating curricula should have some expertise in creating it. Most teachers aren’t getting that training as part of their pre-service or in-service education. I’m not saying that a teacher untrained in curriculum development can’t create great curriculum–maybe they have a gift for it. But, we need to apply the same due diligence to evaluating that curricula as we do other curricula that we deploy. Educational publishers employ trained instructional designers who use models like that of Wiggins and McTighe (Understanding by Design) to draw on the research of best practices, apply proven design principles, and test the results for efficacy.

But the next step is the real innovation: the educational equivalent of an air traffic control system. Each student’s daily exit quiz is fed into an algorithm, which produces the next day’s schedule for each student and teacher. (Teachers get a preview, and can override the schedule.) If a student has mastered a skill, on to the next one. If not, she gets another day’s instruction, this time through a different modality. (The algorithm is aware of which modalities work best for her.) It’s an enormous departure from traditional teaching.

It’s sad that this “adaptive branching” model is deemed “an enormous departure from traditional teaching.” Is it really possible to argue that education should be anything other than basing what we teach next on how what we taught last was understood? The challenge in doing this in a one teacher/30 students model has been the problem, one that technology is well suited to address.

One concern often raised in education is that of testing too much. This is an interesting concern, as there’s research showing that shorter, more frequent assessment leads to better learning. Technology offers the best of both worlds, as the assessment disappears into the background while being constantly there. The line between instruction and assessment becomes blurred just as the student becomes more engaged.

The information we give teachers is either too coarse or too fine. It’s either ‘Johnny can’t do math’ or ‘here are 186 characteristics on seven dimensions of proficiency’ and the teacher says ‘I can’t look at all that.’ The value proposition of School of One is: by telling you what the next instructional step is, we’ll help you thread between that.“

Presumably, such an advance would be well-received by teachers.

Mitchell estimated that School of One costs some $40,000 a year more than other math tech programs, or about $100 per student.

Interesting. I wonder why this is. True cost? Amortization of early R&D? Or margin that the “publisher” is able to extract due to the program’s real or perceived quality?

[In the second year, School of One] added eight more schools, and the collective gains for the 15 schools were 47 percent higher than the national average. They were also highest among the worst-off students, which is exactly what you want to see. … Why did schools do better the second year? New Classrooms is always changing the program, and might have improved something important. Also, everything in teaching improves with practice. … But the gains may be in part due to the fact that these are schools that decided to put a lot of time and money into improving their math teaching,” he said. These same programs usually show less impressive results when tested in a randomized controlled trial that compares math gains in schools that use the program to gains in others that want it but will get it later.

The gains are the key, of course. A year two implementation is early on, and as the article points out, improvement will come with practice. It would be interesting to know the extent to which the program is deployed with the fidelity intended by the publisher.

Why should a school try an expensive, disruptive high-tech platform that’s still unproven? The answer is: in order to prove it. School of One takes comprehensive advantage of technology in ways that let teachers concentrate on teaching. That’s worth getting right. There may be ways to make it cheaper and more effective, but only through further experimentation. As for being disruptive, does anyone defend the current system? “We’re not aspiring to create the least disruptive program,” said Rose. “Our goal is a model that works.”

Well said!