How can we be so confident about our own autonomy if we have not worked out the details of our theory of autonomy? More to the point, in the case of people with addictions, compulsions, and even delusions, how can we be sure that they lack autonomy if we haven't worked out our theory of autonomy? The answer must be that there are broad features of autonomous action that we can identify even if we haven't worked out the theoretical details. We can tell if a car is working well or broken down even if we don't know how the engine works.
Agents can be deprived of their autonomy by brainwashing, depression, anxiety, fatigue; they can succumb to compulsions and addictions. To what, exactly, are we calling attention when we say that, under these conditions, an agent does not govern herself, even if she acts as she does because she thinks she has sufficient reason to do so, even if she has (thorougly) considered the pros and cons of her options, and has endorsed her behavior on this basis, and even if she would have acted differently if there had been stronger reason to do so? Most agents who are capable of asking this question are confident that they are the authors of most of their actions, and are thus accountable for what they do. Nonetheless, as this brief survey indicates, the self-relation they thereby attribute to themselves is extremely difficult to pin down.
But aren't the broad features all we need then for a satisfactory theory of autonomy for it to make the distinctions we need it to make, at least with regard to working out who is autonomous and who is not? Do we need to sort out the details of the debates between coherentists and externalists, or how agents authorize their desires, it sorting this out does not help us make the distinctions we want to make? Even further, can't we conclude that whatever these debates achieve, they don't really tell us more about what autonomy is. We might use the car example: we can understand the concept of a functioning car without knowing how the engine works, and furthermore, knowing how the engine works does not add anything to our concept of a functioning car. To be sure, it is useful for other purposes, but not in the basic use of the concept of functioning car. So with autonomy, the sophisticated debates about self-relations are interesting in their own terms, but they don't tell us more about what we mean by autonomy.
I'm not sure I accept this conclusion, but it certainly is tempting.