I'm not talking about the political view. I'm talking metaphysics here. For those more familiar with theology than philosophy, I'm talking about the view Arminians assume about free will. I just finished a post at OrangePhilosophy about the basic problem, but I'd like to expand it a little bit here mostly because a lot of people who read this blog aren't as familiar with the basic background readers of OrangePhilosophy will normally have.
The libertarian view can be expressed in two non-equivalent ways (and some people hold one and not the other, in which case I don't know if I would call them libertarians).
1. Your action is free only if you could have done otherwise than you actually did. It has to be genuinely possible for you to have done otherwise. If determinism is true, then (on most views) this condition fails.
2. Your action is free if it's caused by you and not by prior events. This condition by definition rules out freedom if determinism is true.
The first principle is called the Principle of Alternative Possibilities (PAP). Harry Frankfurt famously argued against this principle (I think successfully) but still thinks you can meet condition 2 without having alternative possibilities, so he considers himself a libertarian. He just only adopts the second condition. There are compatibilists who accept 1 and give a complex account of possibility to explain how we can be predetermined and still possibly do otherwise. So not everyone who accepts 1 accepts 2, and not everyone who accepts 2 accepts 1. Still, I think 2 is essential for libertarianism, whether 1 accompanies it or not. Therefore, I'm going to argue against 2, which is commonly called the concept of agent causation.
Peter van Inwagen is probably one of the most prominent defenders of a libertarian view nowadays. His Metaphysics book has a chapter on freedom that I think is the best place to get a less technical overview of the current debate. (I use it when I teach free will in a 100-level class, but there's enough detail there to serve for an introductory mixed graduate and undergraduate metaphysics class). Interestingly, van Inwagen agrees with what I'm about to say critiquing the notion of agent causation and just accepts that free will is a mystery but a better mystery to accept than the alternatives. That's where I disagree with him, but we'll come to that shortly.
[Note: from here on, much of this post is an expansion of the aforementioned OrangePhilosophy post.]
The real problem goes back to the issue of how indeterminism can lead to free will. Lucretius infamously explained freedom in terms of a swerving atom. Epicureans were atomists, thinking physical atoms make up the fundamental structure of the universe. Given that, where is freedom? Atoms obey physical laws and conform to deterministic behavior. Lucretius brings in the swerve to explain indeterminism. Atoms will just swerve off their seemingly law-guided path for no reason at all. The problem is that, while it explains indeterminism, it doesn't explain freedom. How can a random event that causes me to do something explain why I'm free? It seems to do the reverse. If a student's hand shoots up in class after I ask a question, I'm going to assume the student wants to answer the question. If the hand-raising was caused by a random event, however, the student doesn't seem at all responsible.
So indeterminacy alone doesn't explain freedom. Something totally out of anyone's control can't explain why I'm free. That then leads me to ask some questions about the choice that the libertarian claims is free in the libertarian way. Is it caused? If it's totally uncaused, then I didn't have anything to do with it. If it's caused, it better have been caused by some event in my control. Call that event A. Is A caused or uncaused? It better be caused, or else it's not in my control. In fact, it better be caused by something in my control, or it won't be in my control. Then we need a previous event B that caused it that's in my control, and you can quickly see that B will need to have been caused by C, which also must have been in my control and therefore caused by D, which also... We get an infinite regress of causes. This would be required for every free choice I make. (This doesn't require that it go back infinitely in the past, since a series with no beginning can all be after a certain point. It goes back asymptotically, getting closer and closer to a moment in time but never reaching it, each event being, say, halfway between the asymptote and the next event. It's also not a contradiction. It's just really, really weird.)
The standard response from libertarians is to deny the whole setup. My choice wasn't caused by a prior event, but it wasn't uncaused either. It was caused by me. I wasn't caused to cause it, but it wasn't random either. Somehow, mysteriously, my will is uncaused to cause it but is still operating in my control. That's how agent causation is supposed to go.
There's still a problem, though. What about the event of my causing it? The event of the choice was caused by me. We still haven't figured out whether the event of its being caused by me is caused. In other words, my choice is caused by me, but my causing of my choice is an additional event that needs to be explained. If it's uncaused or caused by something random, I'm not free. If it's caused by prior events, we have the same problem as above. If it's caused by me, then we have to ask whether that causing is also caused by me and whether that causing is also caused be me and so on. Another infinite regress arises.
So I think the libertarian has three options:
1. Concede to infinite regress in the first argument and just say that every free choice has an infinite number of preceding events all in my control, none of which has any explanation why it's in my control except for previous ones. This fails to explain why any should be in my control. (This must be asymptotic with no first event but all of which are after a certain point, or else they'd go back before I existed.)
2. Concede to infinite regress in the second argument and say that every choice caused by me is an event also caused by me, which means there is an infinite series of causings by me for every choice I make. This fails to explain why any of those causings is caused by me. (This must be asymptotic with no first event but all of which are after a certain point, or else they'd go back before I existed.)
3. Refuse to admit that there's any event of my causing it. I cause it, but that's not an event. I'm not sure why there would be any motivation for such a view, so it seems terribly ad hoc. Furthermore, there seems to be motivation to resist it. How can something happen without there being an event of its happening? If I cause something, there should be an event of my causing it. I just don't see why there wouldn't be.
When van Inwagen considers all this, he declares free will to be a mystery. He says compatibilism has a worse mystery, that we can be predetermined and yet be free. Hard determinism (the view that we're determined and not free) also has a worse mystery, that we can fail to be responsible for our choices (we can hold each other responsible, but there's no moral significance to it). So he accepts the libertarian mystery. My own intuitions are far more willing to accept the compatibilist mystery, since that one at least explains why we'd be responsible if we're predetermined. Our choices are caused by previous events, but they're the right kinds of events. They're events in our own minds, events having to do with our beliefs, desires, values, and character. We act based on who we are. Philosophers as antagonistic toward each other's views as Jonathan Edwards and David Hume (not that they knew of each other) agree on this, and I think they're both right.