Yes!
One of the greatest failures of current science education is weaving knowledge integration. We're doing a bit better than Feynman's description of then Brasil, with grad students having memorized an encyclopedia of science factoids, but utterly unable to reason with them. But even now, ask first-tier astronomy grad students "a 5-year old asks, What color is the Sun ball?" and "What color is sunlight?", and it's usual to get a wrong answer to the first, and not uncommon to get a correct answer to the second, sometimes followed a "<pause> that doesn't make sense, does it?", as two incompatible factoids seemingly collide for the first time since Kindergarten. And first-tier physical-sciences grad students' grasp of color can get really wonky. And then there's everyone else. So we're maybe doing better than Feynman's Brasil, but the available room for improvement is not small.
Anyway, interdisciplinary richly-interconnected sandbox/playground play seems it might help?
And I've seen little like that. Years ago, there was a computer game (name escapes me), loaded with a big database of objects. So users, IIRC, could request almost arbitrary "I want to wield a <thing>, while riding a <thing>". IIRC, riding dinosaurs was popular. So imagine using that to informally teach about the world. "I'm sorry, your giant ant collapsed upon creation, unable to support it's L^3-scaled body on L^2-scaled leg thickness. Give it elephant legs instead?"
> A virtual learning environment
And it needn't be entirely virtual. Witness projective sandboxes.[1] I recently saw a talk by Cynthia Breazeal, where a furry robot, mounted on a desk next to a tablet with integrated apps, provides peer mentoring and emotional support. So even if developing visual systems constrain the use of AR glasses to "a shallow-3D virtual screen", we might still leverage mixed reality.
[1] https://arsandbox.ucdavis.edu/