It is interesting to attempt to specify necessary and sufficient conditions for an action to be free. Incompatibilists (hard determinists and libertarians) hold that one necessary condition of an action's being free is that it is not (completely) caused. This clearly does not provide a sufficient condition, since various subatomic events (e.g. the decay of a radioactive atom on a particular date) are not completely caused, but are also not therefore free.
Compatibilists think that in fact not being caused is not only not a sufficient condition for freedom, but not even a necessary condition. The basic trouble with the idea that free actions must be uncaused, in their view, is that, to the extent that an action is uncaused, it seems to be random: we happen to perform one action, but there is no explanation of why we performed this action instead of some other. This seems to reduce the realm of free action to utterly trivial decisions: if I am completely indifferent as between chocolate and strawberry, then my decision between the two flavors might be random or arbitrary in the relevant sense. But the cases in which we are most interested in freedom of action are cases in which we do have reasons that favor one action over another. It seems that in such cases there is an explanation of why we perform the action in question instead of some other action, and the fact that there is an explanation of the action seems to imply that it is caused. The view of compatibilists, then, is that free actions are not uncaused actions, but rather actions that are caused in a particular way.
A first attempt at a compatibilist definition might be this:
1. An action is free if and only if its cause is internal to the agent rather than external to the agent.
However, this clearly is not a successful definition, since an action can have internal causes and yet not be free (for example, sneezing has internal causes but is not a free action). A second attempt might...