Compatibilist philosophers often suggest that you act freely whenever your actions stem in the right sort of way from your own psychological states. (For example, they may say that you act freely to the extent that you act on your own values, or to the extent that you engage in the right sort of deliberation.) Of course, your psychological states might themselves be determined by your genes and environment, but compatibilists often say that the causal origin of your psychological states is neither here nor there. As long as your actions arise in the right way from your own psychology, those actions are free.
But now suppose we introduce a somewhat fanciful thought experiment. Suppose that a skillful manipulator has set things up from the very beginning of your life to make sure that you perform a particular action. He carefully arranged your whole childhood, exposing you to certain friends, certain TV shows, etc. In the end, his plan works perfectly. You end up acquiring exactly the beliefs, desires and values he wanted you to have, and as a result of having those psychological states, you perform exactly the action he was trying to get you to perform.
Now we face a problem. According to many compatibilist theories, since your actions stemmed directly from your own values, your action should count as perfectly free. Yet it seems intuitive that your action is not free at all. So it seems intuitive that an action can stem directly from your own values but nonetheless be unfree. This is the famous manipulation argument, and it is widely thought to be a big problem for certain compatibilist theories.
In a new paper (forthcoming in Philosophy and Phenomenological Research), Chandra Sripada takes on this argument and reports some intriguing new experimental data.
Sripada shows that people do indeed see the manipulated agent as unfree, but he also shows something more surprising. It turns out that people do not think that the manipulated agent's action is truly expressing his own beliefs, desires and values. Instead, they seem to be drawn to a more complex picture. They think that the manipulated agent does endorse his action on a superficial level but that, deep down, there is a part of him that completely rejects what he is doing.
So maybe the case isn't actually a counter-example to compatibilism after all. The reason people don't see the agent as acting freely is that think the action didn't stem from the agent's own deepest self.
[I'd love to hear what people think of this argument, so feel free to write in, even if you haven't gotten a chance to take a look at the actual paper.]