Most of the reasonable definitions I've heard of for what it would mean to have 'choice' can work with Compatibilism, so I don't consider it a cop out. What definition of free will and 'choice' are you using that causes the problem?

A algorithm can choose something (e.g. netflix will choose a particular film to recommend), and it doesn't cause anyone to worry about free will.

My favourite way of thinking about 'free will', is that it is what it feels like from the inside of an algorithm that is choosing.

I sort of buy this. Anyone that doesn't should put forward a definition of free will that doesn't contradict itself.

The problem is not merely ontology, but the definition of free will itself. The Stanford Encyclopedia of Philosophy has a reasonably long page on compatibilism. "Reasons-Responsive Compatibilism" is an interesting take on the subject which is worth familiarizing yourself with, if you are interested in the subject.

http://plato.stanford.edu/entries/compatibilism/#ReaResCom

I had this revelation when I was 10. Then about 6 months later I realized that the "grand equation" (didn't know about state machines at that age) was irrelevant as I would never be able to see/know all of it (at least in my lifetime).

Consequently, people have "effective" free will - because the knowledge space "to know all outcomes" is too vast to control/comprehend completely.

Even if every action is purely a function of the gears turning, you still have state. As soon as you have state (memories/mood/etc), you have some level of free will in the immediate sense.

Event X happens. How you react depends on your state of mind.

Same problem as with coding. As soon as outputs are not wholly dependent on inputs, and they can have side-effects, all bets are off. Anything could happen.

But your state is also deterministically a function of your prior experiences, per this model. If you consider the input the entire stream of events, rather than each individual event, the output stream will be deterministic even with state.

So basically it's still a 'pure' function in the end?

Per the materialist/deterministic 'there is no free will' mental model, yes: the function taking the input stream is pure.

On the per-event basis, this model boils down to a 'function' evaluating each event with multiple outputs and inputs:

Inputs: the stimuli, mental state prior to stimuli

Outputs: your actions and resultant mental state

The output mental state is then the input for the next iteration/evaluation.

Internal state makes understanding the output more difficult: your output is now a function of the input plus the internal state. However, it does not make it freer: the same input paired with the same internal state will always produce the same output.

But if your actions still could change randomly with your personality, memories and everything else that defines you how would that allow for free will?

Say the universe is perfectly deterministic. You could look into the future by simulating everything perfectly but why would that prohibit free will? Most people wouldn't argue that an all knowing god would mean free will is impossible, why is this different?

I think randomness would undermine the "will" part of "free will", whilst perfect determinism would undermine the "free" part of "free will".

> Most people wouldn't argue that an all knowing god would mean free will is impossible, why is this different?

I'd argue that an all-knowing God is something that can't really be reasoned about. It's the introduction of an infinite value into the discussion (or maybe more like a division by zero?). You end up with logically contradictory statements when describing the properties of such a being.

But isn't it essentially just taking away time as a factor which is pretty much what predicting via determinism would do?

Maybe the general definition of free will most people seem to go with just doesn't deal with weird edge cases because it basically developed to fit reality.