I agree. I am excited to see what type of real world problems we can generate across the science disciplines and grade levels!
I think you would have to start with some basic foundational knowledge of whatever science topic you are studying. You could do this in whatever way you normally teach. Then you could use a computer model to gather some data on the thing you are studying. You could then “look under the hood” and examine the code to see how the model works. You could do an example and then ask students to work in pairs or groups to annotate the code to identify what the agents are, the rules they are following, etc. As a class, you could come together to share these understandings of the code, and try modifying them, making predictions to test what you think will happen after modifying a variable. Groups of students could identify (or choose from a class-generated list) variables to modify and could gather new data, reporting out to the class. Finally, students would learn how to code and would create models of their own. I would imagine working in pairs would be best as both partners could contribute while still having interaction.
A successful science classroom looks like the teacher being the facilitator and students working together in groups on a experiment that they designed based on a problem that they feel invested in. Computers will be used to create their model and run the simulation. Once final results are in, students present their findings to each other where they can get feedback.