GTMOOGLE - you open up a whole new can of worms here! Right on the spot. Control is also one of my major research interests as it is the effect of learning to use the consequences of spontaneous behavior (i.e., operant learning).
Once operant learning ha evolved, feeling in control evolved to be rewarding by itself, even in flies: if flies are allowed to chose between a situation of full control or reduced control, they choose the full control version.
Obviously, there's a lot of research going on in this respect and "free will" in my opinion is a much too narrow term to fit all that in, even though some of that factors into what we commonly call or experience as free will.

ruzel, the behavior of nonlinear systems is very difficult to predict, but it's not random. There is a gradient in predictability where "random" and "determined" are just the endpoints.
So nonrandom spontaneity is not an oxymoron.

gtmoogle, in our research article, we have analyzed the spontaneous behavior of flies in three different environments and have found that despite the behavior being very different to the human eye, the same nonlinear signature could be detected. So quantitatively, the behavior is different in different environments, but qualitatively it remains very similar. This is reminiscent of what people see when they image the human brain using fMRI: spontaneous activity during idle time, which changes (but is still there) when we engage in a task.
I'm sure our evaluation algorithms will be used on other animals in the future as well, not only on different generations of flies (which we will have to do anyway, when we look for the brain areas which generate spontaneous behavior).

I'm the senior author of the study and Matt was one of my students in a course on neuroinformatics this past term.
There are some very good points raised in the comments so far, especially the ones by gtmoogle and slida.
We were very careful not to raise the issue of free will in the original research article. There are currently three different levels at which "free will" is discussed.
(1) One is the dualistic free will in which our will exists independently of our brains. This, of course is an outdated view. Our choices are made by us, i.e. our brains and hardly any neurobiologist still holds a dualistic view any more. So that's not what we're talking about.
(2) The second is our own subjective experience of free will. Why do we have this strong feeling of authorship and how does the brain generate this feeling? For obvious reasons, that's not the level we are talking at, either.
(3) The reason for our subjective feeling of free will and authorship: is the brain just a complicated robot that works according to fixed rules and any variability is simply just noise as in a radio tuned between stations? Or is the variability that we see actively generated by the brain and not just some random by-product of inevitable noise? This is the level we address with our study. What we find is that already fruit flies have evolved brains which actually appear to generate more variability than they would if their brains had evolved to be as precise as possible. In other words, already flies are less precise and more fuzzy than they would technically need to be, from a technical point of view. Biological organisms are not engineered. There are many reasons why this has evolutionary advantages. Probably one of the easiest ones to cite would be, that it is more difficult for predators to catch the fly, if it makes unpredictable turns and at the same time, the fly can still head for a safe place and is not just flying randomly anywhere. There's a third way between deterministic and random and that's the one the flies are taking.

The most straightforward implication of our mathematical results is that individual fly behavior will never, in principle, be fully predictable, only probabilistically. Similar to quantum mechanics, individual behavior is fundamentally indeterminate and the reason is not chance, but the inherent property of brains to always do the same thing slightly differently.
IMHO, this is as close as biologists can come to scientifically address the old question of free will: how can the brain produce different behavior under identical circumstances? Our data show that even the simple brains of flies can have several behavioral options in the same situation - so I think it is likely that humans also always have a choice (as long as they have intact brains).

The distinction between "random" and "probabilistic" is a very fine one, but it decides how research into the biological workings of choice will progress. As such, our study showed us in what direction we need to focus our attention when we now start to investigate the brain areas generating variable behavior.

Just last Friday a camera team from JoVE was in my and Alexander Maye's (one of the collaborators) lab to shoot the technical details of our study. With the JoVE publication, we will release our entire set of raw data as well as the source code for the mathematical evaluation free for anybody to reproduce our results. So keep your eyes on JoVE for the full and open scoop of this research.