Seth Lloyd has posted a well-written pre-print, proposing a self-administered Turing test for free will but also dealing with some other aspects of the free will debate. Some excerpts:

... the theory of computation implies that even when our decisions arise from a completely deterministic decision-making process, the outcomes of that process can be intrinsically unpredictable, even to – especially to – ourselves. I argue that this intrinsic computational unpredictability of the decision making process is what give rise to our impression that we possess free will.

It is important to note that satisifying the criteria for assigning oneself free will does not imply that one possesses consciousness. Having the capacity for self-reference is a far cry from full self-consciousness...An entity that possesses free will need not be conscious in any human sense of the word.

This paper investigated the role of physical law in problems of free will. I reviewed the argument that the mere introduction of probabilistic behavior through, e.g., quantum mechanics, does not resolve the debate between compatibilists and incompatibilists. By contrast, ideas from computer science such as uncomputability and computational complexity do cast light on a central feature of free will – the inability of deciders to predict their decisions before they have gone through the decision making process. I sketched proofs of the following results. The halting problem implies that we can not even predict in general whether we will arrive at a decision, let alone what the decision will be. If she is part of the universe, Laplace’s demon must fail to predict her own actions. The computational complexity analogue of the halting problem shows that to simulate the decision making process is strictly harder than simply making the decision. If one is a compatibilist, one can regard these results as justifying a central feature of free will. If one is an incompatibilist, one can take them to explain free will’s central illusion that our decisions are not determined beforehand. In either case, it is more efficient to be oneself than to simulate oneself.

Q1: Am I a decider?

Q2: Do I make my decisions using recursive reasoning?

Q3: Can I model and simulate – at least partially – my own behavior and that of other deciders?

Q4: Can I predict my own decisions beforehand?