Kaj Sotala, you make a good analysis about the meaning of concrete particular actions in the game: if they are connected to the bigger narrative meaning of the game and have dependencies to the future, then the actions make sense (Salen & Zimmerman). You have the game world, why wouldn't you make particular actions meaningful and related to it (well, maybe because you are just gluing game elements on top of a traditional learning problem or activity).
I'd like to take up on an issue though: The question of transfer is an empirical one or needs more references than you wrote. You seem to assume transfer would be harder in “stealth learning” (or implicit abstract learning from concrete, contextual, and situational experiences), because (again assuming) a contextualized learning wouldn't encourage reflection. But does reflection in fact encourage transfer or not? This sounds really basic and my bet would also be yes, but I think we'd be better of with some (reference to empirical) data here ;)
I general, I'd think the question of abstraction and transfer is interesting here – especially after gone recently through a multidisciplinary study project with people that were not familiar with analytic and logical thinking nor clearly defined concepts (i.e. scientific thinking), http://mainiosocial.com
My question is, are you better off when learning from i) a contextualized and concrete situational experience/example or from ii) an abstract and generalized "academic" (logical, mathematical, etc.) representation? What's the "better off", you might ask? I'd define it in terms of transfer, that is, in how many future context the learner is able to apply this thinking to understand situations and make (rational approaching) decisions? We can now ask, what do we actually mean by abstract here, along the previous functional style thinking and definition?
I'd like to think, applying my work-in-progress bachelor's thesis, that "what is abstract" depends on a personal learning history. That is, for example, if you have invested your time heavily on logic calculations, you are in fact made logic very concrete to yourself. Yet, learners differ. Some might not see the practical relevance of logic in everyday life, whereas for some it might be easy to apply.
The additional benefit in this style of "abstract learning context" is, that the abstract here includes an explicit assumption of "being applicable to numerous contexts". Does that happen, depends on the learner also.
The crucial metric here is the transferability, the cross-situational individual future applicability of learned material. On the general level, I consider the source representation secondary; but we surely find learner differences in what kind of material is the most easy to digest and apply.
Note: I don't talk about abstraction or abstract thinking here, as I rather use its subjective definition above. I find it more beneficial to talk functionally about "transfer ability", that is, how to elaborately apply what was learned. Abstraction is a technical, intermediate, and subjective term, whereas transfer is concerned with in-situ abilities and behavior, which is more concrete and interesting for a broader audience.