It is part of the standard explanation that determinists offer, e.g. Wegner, The Illusion of Conscious Will. It has problems, however. Error has to be established before it is "explained". Free will, like qualia, consciousness, or determinism itself are formulated in a way that makes them empirically undecidable, so the lack of proof is moot. And we posit entities to explain all sorts of observed behavior, habit breaking isn't special. That in itself tells us little about the veracity of what is posited

If you use the idea of conditioning (rather than habit) then you're speaking of a well-understood phenomenon. Gurdjieff calls most human beings robots since their behaviour is conditioned and not free, and this would be the standard view of behaviour for his Sufism and mysticism in general. Freedom would consist in freeing oneself from conditioning, which means freeing oneself from the conditioned self. .

@Conifold Not habit breaking as such. Rather the inconsistency in following conditioned behavior, as if there exist some power to decide whether a habitual action gets performed or not.

@PeterJ Will look into that. Interesting though that most would take habituation as a negative wrt freedom. Is there anyone who forwards a positive view of conditioning, e.g. I drive my car almost automatically leaving my thoughts free?

@christo183 - Conditioning of that sort (driving) is useful and might be viewed as a positive skill. It is this conditioning we are paying for when we take driving lessons. It would all depend on whether the tail is wagging the dog or vice versa.

I think you're conflating analytic dimensions. A lot of behavior is habituated. That's necessary: we would not be able to walk if we had to consciously control the interactions of all the muscles that act required. But we still have the experience of a subjective mind that decides (say) where we are going to walk to. the subjective understanding that we can change our minds is what leads us to believe in free will. Habituation will move us along in any direction we point ourselves, but we (ostensibly) reserve the right to choose the direction we point ourselves in.

@PeterJ if we were dogs we'd be conscious of being wagged or of doing the wagging. The point is that it is not always one or the other (and we are aware of that). It is this inconsistency that I'm positing as the experience of free will.

@TedWrigley I'm intrigued as to what "analytic dimensions" you speak of? Also see comment to PeterJ.

Two analytic dimensions: behavior and choice. one's body has to do different things to throw a fast-ball or a slider — that's behavior — but there still must be some point in time where we flick a switch to invoke one behavior or the other.

@Conifold - "Free will, like qualia, consciousness, or determinism itself are formulated in a way that makes them empirically undecidable" If you're talking about non-compatibilist free will, wouldn't that in principle be testable along with determinism, if for example we could design some machine that could correctly predict what any physical system including a human would do? Or slightly more realistically, if mind uploads turn out to be possible, and they behave just like us, wouldn't that be strong evidence we don't have free will either?

@Hypnosifl Such machines are impossible aside from free will, due to quantum indeterminacy, and what is to preclude ascribing free will to whatever the "uploaded mind" turns out to be? The problem is that determinism (or lack thereof) can be made compatible with any empirical behavior, so the question can only be resolved holistically, by selecting a paradigm that provides a more comprehensive and coherent overall picture. As it was resolved in the classical mechanics in favor of determinism, for example, on structural grounds, not by any "testing".

@TedWrigley You're right about choice and behavior, however in habitual behavior it seems we can still override: one may, by choice, override the normal action and walk like a zombie for instance. Without free will, automatic behavior should always be the same and any deviations would require explanation... Though I imagine there can be several such explanations, I was wondering in this question whether anyone has taken this to be evidence of free will.

@Conifold - For a biological human such a machine might be ruled out by QM, but algorithms running on a classical computer are deterministic in the same sense that arithmetical sums are--if you have multiple machines separately implementing the same computation and they don't get the same results, it's due to a malfunction in the machine (whether due to unlikely quantum events in the circuits or other issues). If an upload could be implemented on a classical computer, and it were restricted to sensory input from a virtual environment, one would expect to have this sort of repeatable program.

@Hypnosifl Quantum systems can be simulated on classical computers, and we already have chatbots that humans have hard time distinguishing from humans, and AI that play chess like humans. One means that humans chat and play chess that way no more than the other means that quantum systems are classical. Simulation isn't reproduction. Deterministic AI may well be able to simulate humans one day, by itself it won't tell us how human behavior is actually produced.

@Hypnosifl This was a reasonable expectation (as the Turing test idea shows) before the advance of chatbots. They showed just how little what people "feel" is worth, too little filters into behavior and is picked up on to be conclusive. A credible AI "upload" would have to do more, I suspect, reproduce biological neural networks at some structural level, rather than just their overt behavior. And the result will likely be the loss of understanding, the same uncertainties that surround biological carriers of "free will" will surround artificial ones.

@Conifold - Yes, the idea of an upload is that it'd be a simulation at the level of synapses at least, based on mapping a biological brain at that level of detail. And this may well mean we wouldn't understand its functioning conceptually, but you could still feed the same sensory input to multiple copies to see they behaved identically. As for fooling ppl, you could interact with uploads the same as with anyone else, having long-term friendships etc. (I doubt a chatbot would fool a person if they interacted for years, especially if it was designed to fool the friends of a preexisting person).

@Hypnosifl Neural networks are sensitive to small perturbations of initial conditions, at least in some regimes. Determinism in such systems can only be sustained by assuming infinite precision, which is cut off by QM. They will not behave identically on the same inputs, free will or not. "I feel that it behaves like a person" is mostly of sentimental value, even if chatbots could pull it off for years (maybe one day they will). I can only see a comprehensive package, including that, mapping neural patterns to "thoughts", theory of emergent mental, etc., convincing most people.

@Conifold - An analogue neural network would have that problem, not a simulated neural network on a classical digital computer (a Turing machine), multiple copies of the same digital program given the same inputs and starting conditions should give the same output even if the program is a simulation of a chaotic system like the weather or a brain. And if you don't value the evidence of extended interaction, conversations which exhibit understanding of complex issues and emotions etc., what reason do you have to believe other people have a consciousness analogous to yours?

@Hypnosifl Given the number of synapses simulated, and the types of dynamics involved, roundoff errors will be enough to doubt that it is faithful to its object. One would also have to accept some level of "glitching" from background noise, so multiple programs will not run identically anyway. It is not that I do not value prolonged exposure for practical purposes, but for those purposes what does it matter what the metaphysics is underneath. We anthropomorphize inanimate objects as a matter of course when it is suitable, it is structural resemblance that takes it beyond that.

@Conifold - The point of an upload is to be qualitatively similar to its object, like a simulation of atmospheric dynamics or galaxy formation, not to precisely duplicate how the original brain would behave in the same circumstances down to correctly predicting when every neuron fires (prob. impossible due to chaos theory amplifying quantum randomness). If you think multiple runs of complex digital programs (like the kinds of physical simulations I mentioned) don't behave identically, I think you're misinformed about the error level of modern digital computers, what's your basis for that?

(continued) As for people being fooled, I don't think you should conflate anthropomorphization of objects in ordinary speech or intuition with literal belief that an inanimate object has humanlike consciousness, the former is common but the latter basically never happens except with examples like brief encounters with chatbots that are designed to continually change the subject and avoid giving specific answers to most questions. Also if you think humans would be easily fooled even in long-term relationships, what's your answer to the question about why you think other humans are conscious?