Why simulation, honestly

Across independent studies, interactive simulations match or beat traditional instruction on understanding. The conservative meta-analytic figure is an effect size of roughly 0.6 (the SRI 2014 review of K-12 studies). Some single studies report much larger numbers; we deliberately cite the low end.

The caveat matters: education studies tend to overstate effects, so treat about 0.6 as “what the research found,” not as a promise about any one learner. We never put a learning-gain number on letsgye itself.

A sim on its own is a commodity. PhET, the large free library, describes its simulations as “just one piece of a well-designed curriculum.” A folder of standalone sims is not a sequence, and the evidence is clear that unguided, free exploration actually loses to plain explicit instruction (a negative effect of about -0.4 in Alfieri 2011).

What the evidence credits with the gains is the structure wrapped around the model: a real sequence, scaffolding, mastery gating, predict-then-explain, and getting the learner to explain in their own words. In the studies, assisted, guided discovery helps (about 0.3 in the same Alfieri 2011 review), and eliciting self-explanation helps as well. Scaffolding, worked examples, and feedback are exactly the components that turn discovery from a loss into a gain. Bare sims alone do not reliably build reasoning until that structure is added.

That spine, sequenced zero to expert and mastery-gated, is exactly what letsgye is.

Letting a learner predict and wrestle with a problem before the explanation helps understanding and transfer (about 0.36 in the productive-failure research), but only when the explanation then builds on what the learner actually tried. And it backfires for the youngest learners, who do not yet have the background to generate productive guesses.

So letsgye ladders it by age: the youngest band gets clear instruction first, then the sim; older learners predict first, operate the model, and then we explain against their guess.