I think this definition is a good step towards understanding free will.
In particular, it helps address the question of whether the fact that a
system's behavior is predictable from outside, or simply deterministic,
means that it does not have free will.
However by itself the definition is too simple. There has to be a
specification of some minimum predictive ability, otherwise the definition
is satisfied vacuously by systems simply because they can't predict at
all. Cars, inclined planes, and knives all appear to have free will by
this definition, because they can not predict what they will do in the
future.
One possibility would be simply to add that the system must be concious
before we begin to address whether it has free will. This would still
leave the definition useful for the cases we are interested in, without
having this distracting loophole.
Hal