Perhaps the idea of Molinist type counterfactuals could be aided by the ‘extreme modal realism’ theodicy which I have previously rhetorically appealed to. I am not attracted to extreme modal realism or anything like it, but let us suppose for argument’s sake that there are very many actual worlds, and that to say that ‘x’ is possible really is just to express the fact that ‘x’ is actual in at least one world. Perhaps we could even shore up our extreme modal realist theodicy in such a way that we restrict actual worlds to any and all worlds which it is feasible for God to create. Here I appeal to a notion suggested by William Lane Craig that perhaps there are morally sufficient reasons for God to prefer our world to a world in which all persons are saved, such as that the only such worlds which are feasible for God to create are those in which there are only five or ten people. So, Let us suppose that God has created a sufficient number of worlds to cover over all worlds which it is logically feasible for God to create.
I think maybe this could help us construe Molinist type counterfactuals as counterfactuals which are grounded in the reality of at least one actual world. Thus we could say that all true counterfactuals are only counterfactual relative to our actual world, but are straightforwardly factual and realized elsewhere in the multi-world ensemble (not to be confused with the multiverse, though it is a close ‘Modal’ parody).
However, even on this view of counterfactuals, granting as it does the gratuitous assumption that God created the set of all worlds which it is feasible for God to create, I still think the counterfactuals are only trivially true. We cannot say, if we believe in libertarian free will, that ‘Joshua would choose x over y, if he were in A1 situational circumstance’ is true ‘at’ our world merely by being true ‘in’ another world which is relevantly similar to our own. The counterfactual, it seems, is not true qua counterfactuality.
Therefore, it seems to me that even this construal of counterfactuals would present us with ungrounded counterfactuals. If we say that some counterfactual of the Molinist type is true, we do not mean that it is actual in some world where all the (other) same circumstances exist as they do in our world, but rather we mean ‘were it true that A1, it would then be the case that Joshua would choose x’. There is, it seems, just no way for that counterfactual to be true if one believes in libertarian free will.