Suppose that George voted for a friend to be hired, and now he want to figure out whether he did it for the sake of friendship, or whether he did something nepotistic. One way for George to figure this out is for him to ask:
- Were Jane not my friend, would I still have voted for her?
But, now, I think that
- Were Jane not his friend, George would still have voted for her
When George asks (1), he is looking for an answer that supervenes on facts about his motivations. The Molinist answer to (1) does not do that, though it may be probabilistically connected with facts about his motivations.
At the same time, there are times when we really do want to know the truth of a Molinist conditional. Thus, prior to the vote, Jane might ask herself:
- If I were to cease to be George's friend, would he vote for me?
The above raises a worry for Molinists that they have to have two kinds of subjunctive conditionals of free will, the Molinist and the non-Molinist ones, while anti-Molinists need only one, the non-Molinist one. Maybe, though, the Molinist can say that when Jane is trying to figure out George's motivations and character, she is not interested in the truth value of the B→V conditional (were I to break off, he'd vote for me), but in the probability of that conditional. (See also this post).
Were Jane not his friend, George would still have voted for her is not a Molinist conditional. Here is why. What George wants to know is something about his actual motivations. The truth value of (2) understood Molinistically is irrelevant to how things actually went--there is another possible world, where everything in fact goes just as it does, but where (2) understood Molinistically has a different truth value
That's interesting. Is it irrelevant? Suppose there are probabilistic reasons based on George's character that explain why he did it out of friendship. Those same reasons obtain in worlds where he does not act in character. As with all indeterminsitic events, we have statistical explanations for what George did. Of course there are also worlds in which George has all the same reasons but does not vote for Jane. In those worlds it is inevitable that the CFF conditional backtracks. The objective probability for his voting for Jane cannot be the same, for instance, nor is what God knows the same, or what the perfect predictor predicts (if such there be). But this is good news, since it offers an explanation for what he does in those other worlds without badly affecting libertarian freedom.