William Lane Craig responded to Prof. Anderson's article regarding the "Fallible God of Molinism" in a predictable way - by arguing that although the possible worlds where the people choose otherwise are possible, in those worlds God's knowledge would be different. Thus, although a person could not bring about a world where God's knowledge is wrong, he could bring about a world where God's knowledge is different. That sounds confusing, and raises several objections.
A first layer of objection is that in this allegedly possible world is a world with different circumstances than the actual world immediately prior to the alleged free choice. In the actual world, God already has knowledge that the man will do X. Thus, in this set of circumstances, saying that the man "can avoid doing X" on its face requires different circumstances.
A typical Molinist response can be seen in Dr. Anderson's summary here (link):
Craig is correct that I made a misstep in the original argument, as I acknowledged in reply to a comment by Greg Welty. I said at first it wasn’t relevant whether God’s decree (or “providential plan,” to use Craig’s terminology) is included in C, i.e., the circumstances in which God places the free agent S, knowing that S will do A in C. But that’s mistaken, because Molinists will say that C doesn’t include God’s decree, such that in those possible worlds in which S does not do A in C, while C is the same, God’s decree is different. (As Craig puts it, God’s providential plan is not “firm and fixed” across worlds in which S chooses in C, even though C is fixed.) The counterfactuals are different in those alternate worlds and therefore God’s decree will also be different in those worlds, since it is based on those counterfactuals.In other words, people like Craig argue that because God's knowledge is not the cause of the future action of the person, it can properly be excluded from the circumstances. Anderson mentions how this looks like special pleading. Molinists typically argue that it should be clear that God's knowledge is not itself the cause of the future event. Presumably, Molinists would ground God's knowledge in his decree to instantiate and his middle knowledge.
Nevertheless, in some cases God's knowledge is plainly one of the causes of human action. For example, we have a category we call "self-fulfilling prophecies." While this category is not usually applied to God, we note that God's knowledge is plainly the basis for actual prophecies. Furthermore, some people whose actions are prophesied hear the prophecy and act based on their own knowledge of the prophecy. For example, leaders of the Israelites were sometimes motivated to go up to battle by prophetic revelation of victory in that battle. In such a case, saying that God's knowledge is not part of the relevant circumstances seems totally indefensible.
Worse than that, Molinists are overlooking the problem that even if God's knowledge itself were always kept secret and were never revealed, there is an antecedent cause of God's knowledge, which is also a cause of all the (other) circumstances. Anderson explains it this way:
The Molinist has to say, in effect, “Don’t worry, it’s possible for S not to do A in C provided we don’t include God’s decree in C, for then it would be possible for S to do other than what God has decreed.” In other words, the Molinist has to restrict C to intramundane circumstances.
But why should the Molinist be permitted to make that move? After all, the circumstances are causally connected to God’s decree (and necessarily so in the Molinist system). The circumstances obtain only because God decided that they should obtain and caused them to obtain. If we can draw the boundary line of C so as to exclude God’s decree and his actions to implement that decree (i.e., his manipulation of intramundane circumstances), why not draw the boundary line even more narrowly, so as to exclude, say, all circumstances more than five years prior to S’s choice, or all circumstances more than fifty miles away from S’s location? Is there any principled reason to draw the boundary line of C where the Molinist wants to draw it other than to save the system? Why think that it being hot and sunny in Charlotte today is a relevant element of my circumstances but God’s making it hot and sunny in Charlotte (partly in order to direct the choices of Charlotteans) is not a relevant element?In other words, God's knowledge is based (at least in part in the Molinist system) on God's decree. But God's decree is always a cause of at least some of the circumstances for every human choice. Thus, it is fair to say, as Anderson did:
[T]here’s a clear sense in which, on the Molinist view, God determines S’s choices. It isn’t a causal determinism, but it’s still determinism in the sense that God’s decree that S will do A (which is fixed prior to any of S’s choices) guarantees that S will do A. Given that God has decreed that S will do A, S cannot do otherwise than A.In short, in order to make the God of Molinism infallible, Molinists have to compromise LFW of creatures by artificially excluding things like the divine decree. But about the only reason to even consider Molinism is an a priori commitment to LFW.
Furthermore, if Molinists are willing to compromise by excluding some of the causal factors of choices, what's the principled basis for rejecting a Calvinistic free will, in which a will is free in the sense that it could choose differently in different circumstances?