My father told me once about a chemistry experiment when he was in college. I can't remember the exact details, but it involved naptha, an explosion, a fire, a cut hand, and very nearly got him expelled from school.
When I was a kid, we visited the CS labs at UIUC where I got to see PLATO in action. The demonstration I saw involved a simulated a bunsen burner, and a flask filled with naptha. The idea was to melt the naptha without causing it to explode. I spent about thirty minutes playing with it, causing about 50 virtual explosions before figuring out how to avoid them.
I guess my point here is that there are several real world experiences that are instructive, but carry significant risk. Modeling them digitally is a great way to introduce concepts with less risk.
Also as a teen, I got to go see several presentations by Papert, and eventually got a LOGO system myself. In retrospect, I think what I learned from both LOGO and Smalltalk was a mode of asking questions about real world object's behaviors.
One project I worked on in high school was to use LOGO to model an atom. It's one thing to sit in class and listen to lectures, and even to do experiments, but it wasn't until I tried to model a Hydrogen atom in LOGO that I learned how much I didn't know about chemistry, physics, and math. That investigation led me to pursue independent research on the subject in an effort to refine my model. I don't think I would have been motivated to do the extra research were it not for the habit of inquiry I developed with using LOGO (and later Smalltalk.)
I guess I owe a debt of gratitude to my parents who introduced me to GOOD computing projects at a young age.