Well, without having read any of the rest of it, the bit you've cited is clearly a joke (and a pretty funny one).
I think he largely obfuscates around the point, as many philosophers often do. He never gives, in my opinion, a satisfactory definition of free will, and even intentionally avoids doing so. After about reading 3/4ths of it I was so frustrated trying to figure out what he meant by the words he was using I just gave up.
To me, the question of free will is the question of a deterministic universe. There are three questions I ask anyone who attempts to argue for free will. 1. Do you believe our actions are controlled by our brain? 2. Do you believe our brains are made of atoms? 3. Do you believe those atoms follow the laws of physics, even if we may not know exactly what those laws are (yet)? It seems absurd to answer "no" to any of those questions to me. But at least at this point I've gotten the person with whom I am arguing to admit that that is eir answer.
I agree that if you don't have a definition of "free will", you're blowing smoke. But there are widely used definitions of the term that are compatible with a deterministic universe. Compatibilism is a long-standing and widely held belief.
For example, a standard philosophical definition is that an actor has free will with respect to an action if the actor's choices caused the action or could have prevented it. Whether, in turn, that choice was determined by the past is irrelevant from the point of view of this definition.
And once you adopt this definition, it can do useful work. In particular, knowing whether an action was the result of a choice tells you whether you're justified in drawing inferences about the actor's mental state based on the action. If you choose to hit me, I can infer things from the action that I could not infer if you hit me while you were having a seizure.
That may be helpful in assigning fault, etc. But that wasn't the question being asked. It's not useful in answering the question of do we have free will? It assumes the answer to that question is yes, by the very nature that it considers choice.
Does that definition of free will add any further understanding? Using choices to explain free will to me seems no different than using phlogiston to explain fire. You're just adding a meaningless layer of complexity that completely begs the question.
My definition picks out a set of things I want to talk about, and captures what I mean when I use the term free will. If you don't like that definition, what's yours?
I assume you want to define free will in an incompatiblist way. If so, I ask what understanding that definition aids or what explanatory power it has?
I gave a definition in my opening comment:
To me, the question of free will is the question of a deterministic universe.
To quote you, my definition picks out a set of things I want to talk about, and captures what I mean when I use the term free will.
If you want to define free will that way, that's fine. But it doesn't help answer the question of do you actually have it.
I was going to answer -- well, I guess I still will, that free will is hopeless even in an undeterministic universe because either actions have a cause and they're determined or they don't, in which case they're not caused (not even by you) so they're still not 'free' in the sense of chosen -- and then I read in the linked essay that this is a misconception.
Now on to the misconception of the anti-free will camp. I've often heard the argument which says that not only is there no free will, but the very concept of free will is incoherent. Why? Because either our actions are determined by something, or else they're not determined by anything, in which case they're random. In neither case can we ascribe them to "free will."
I don't see yet that this is a misconception... for example, I disagree with this:
For me, the glaring fallacy in the argument lies in the implication Not Determined ⇒ Random. If that was correct, then we couldn't have complexity classes like NP---we could only have BPP. The word "random" means something specific: it means you have a probability distribution over the possible choices.
Actually, the word "random" means a few things. One of these meanings is "not determined" (not caused, spontaneous). So Not Determined --> Random is definitional. And then the argument makes sense to me. (If it's not determined, then it's not determined by you.)
I think Scott's purposes here are pedagogical...he is purposely raising issues to be thought-provoking, even if some of the arguments are in the end fallacious. I think most of the points he brings up in that lecture (and others) can be resolved cleanly with some thought, but actually putting the thought in is useful.
But yeah - that sort of methodology doesn't look like it's ever going to resolve the problem. You've got to either give free will definite properties or dissolve it.
This bit misses the point about predictability of goal-optimizing outcome without predictability of individual steps used to reach that outcome:
Now let's get back to the earlier question of how powerful a computer the Predictor has. Here's you, and here's the Predictor's computer. Now, you could base your decision to pick one or two boxes on anything you want. You could just dredge up some childhood memory and count the letters in the name of your first-grade teacher or something and based on that, choose whether to take one or two boxes. In order to make its prediction, therefore, the Predictor has to know absolutely everything about you. It's not possible to state a priori what aspects of you are going to be relevant in making the decision. To me, that seems to indicate that the Predictor has to solve what one might call a "you-complete" problem. In other words, it seems the Predictor needs to run a simulation of you that's so accurate it would essentially bring into existence another copy of you.
I got the impression his essay isn't about whether there is or isn't free will. I think he's just using the question to develop some peripheral ideas. For example, I couldn't tell from his essay whether he believed in free will or not. I don't suppose he considered the answer to the free will question to be immediately relevant.
Given these three assumptions, the theorem concludes that there exists an experiment---namely, the standard Bell experiment---whose outcomes are also not predetermined by the history of the universe. Why is this true? Basically, because supposing that the two outcomes were predetermined by the history of the universe, you could get a local hidden-variable model, in contradiction to Bell's Theorem. You can think of this theorem as a slight generalization of Bell's Theorem: one that rules out not only local hidden-variable theories, but also hidden-variable theories that obey the postulates of special relativity. Even if there were some non-local communication between Alice and Bob in their different galaxies, as long as there are two reference frames such that Alice measures first in one and Bob measures first in the other, you can get the same inequality. The measurement outcomes can't have been determined in advance, even probabilistically; the universe must "make them up on the fly" after seeing how Alice and Bob set their detectors. I wrote a review of Steven Wolfram's book a while ago where I mentioned this, as a basic consequence of Bell's Theorem that ruled out the sort of deterministic model of physics that Wolfram was trying to construct.
That seems to be a silly argument. The outcome is determined by the history of the multiverse. We don't know what that history is, but our ignorance is not a form of indeterminism - and it most certainly does not rule out deterministic models of physics.
http://www.scottaaronson.com/democritus/lec18.html
Am I missing something or is the thinking methodology surprisingly sub-par here? Has anyone else read it?