Imagine a deterministic computer simulation of a complex but deterministic universe. This universe has either evolved or had inserted somewhat autonomous agents that strive to “make a living” and so perpetuate themselves. To do this they must achieve what modern complex organisms of our universe do:
Dennett said that the free-will vs. determinism problem that so many people have stems from a model of the universe that omits the actor—as when a dualistic framework is implicitly assumed where volition is taken as arising in some agent not within the universe and thus unable to influence activity within the universe under the deterministic rule that things are as they are (in the universe) because they were as they were (in the universe).
Our articulation also lays bare some points often overlooked by realists who speak of computations that agents can do in principle. Often such computations must model the entire universe in order to make some choice. It is clear that the overriding simulation, which I assume to be finite, cannot include a simulation of itself. Perhaps some universes can support simulations of some sub component that is accurate with suitable caveats, but I can’t think of such cases and our particular universe seems ill-suited for such. Us organisms can and do form imperfect but useful models. We can’t know all and we can’t even use all that we have correctly guessed to choose optimal actions, for lack of compute power.
I do not propose that the above activity necessarily be conscious. Indeed such stuff must go on in animals that we do not consider to be ‘self-aware’. I like Metzinger’s explanation of the illusion of spontaneous choice: There is a causal series of events, such as described above, many of them subconscious. In such a series we are likely to be aware of only the later events of the series and thus feel as if the process were ‘spontaneous’, i.e. free.