Shaun Nichols on Free Will (Among Other Things) on Free Will

MK: _That gives us good predictive power, and it means that the leftover “free will” part might appear as “total randomness.”_

I'm not so sure. Nobody denies that, absent some artificial control or physiological damage, humans have the freedom to "attend" to different objects. I can choose to focus on my foot, or my hand, or that TV over there, so long as I am not being overwhelmed by stimuli so that such a feat is prohibitively difficult.

These sensory gestalt shifts are not deducible from brain state information, however accurate and timely (thought they can probably be instigated by external manipulation).

As we go forward with the large task of knowing ourselves inside and out, it should become increasingly clear that human agency is weighted, limited, real and very much expandable -- its enablers, the brain, reflection, and self-reification.

Micha:
It’s not logically impossible, but it’s practically highly unlikely, given the enormous computing power this would take and the extremely low pay-off for systematically predicting an individual’s actions.

Sure, maybe. It's very hard to predict the future hundreds of years out, of course. It is more plausible that there would be something more targeted -- a machine that scans your brain and tells you whether you have a "worrying" distribution of activity, and gives advice on how to get back to a "normal" state.

We will probably find that happiness levels go up, murders go down and human actions generally change a lot. This probably isn't stark enough to blow anyone's mind, but it is a very fuzzy version of the same thing (treating the atoms in a person's head as causally predictive of their actions). The difference is it wouldn't be the "ultimate demonstration" of that causal predictiveness.

Free will is basically an encapsulation of our uncertainty about an agent's actions, and consequent need to resort to desire/belief account of action (which can be quite predictive). Insofar as human actions will always be uncertain from the atom or neuron or brain scan point of view, we will always have some room for free will.

Another interesting scenario is if brain scans get better than humans about depicting the person's emotional state. That gives us good predictive power, and it means that the leftover "free will" part might appear as "total randomness."

MK,

But what about that sci-fi movie with Tom Cruise!

If we really could take a supercomputer and predict your actions based on atoms, all hell would break loose. We would be like the remote tribe who you show them a camera and it completely blows their freaking mind. The supercomputer is not logically impossible and it could happen someday.

It's not logically impossible, but it's practically highly unlikely, given the enormous computing power this would take and the extremely low pay-off for systematically predicting an individual's actions. Daniel Dennett makes the same sort of argument for the practical (but not logical) impossibility of things like truly convincing brain-in-the-vat scenarios in the beginning of Consciousness explained. The information processing capacity and continual feed-back loops necessary to achieve such a feat would be prohibitively costly and unlikely to ever be as complete or convincing as mere logical possibility would imply.

I would posit that our revulsion against utilitarian baby-killing is an outcome of some ongoing emotional training process which we are always free to do or not do.

Sorry, partly an outcome of training. Partly native (evolutionarily developed).

Oh yeah one more thing. I think our term "free will" is a pragmatic way of interacting with humans for whom our best prediction mechanism is at the level of beliefs/desires/etc. If we really could take a supercomputer and predict your actions based on atoms, all hell would break loose. We would be like the remote tribe who you show them a camera and it completely blows their freaking mind.

The supercomputer is not logically impossible and it could happen someday.

The moral perspective is interesting here. If we could predict people's actions precisely, does it change morality? Well yeah it does. Game-theoretically, uncertainty has just been radically reduced. The optimal "play" now involves locking people up who are about to do something bad. Or, showing them the supercomputer results so they know they are at risk and can take some classes and otherwise "shape up" appropriately.

We already try to do this. If you commit enough crimes, we predict you will commit more so we lock you up for life. If you get a DUI, we predict you will drink more in the future so we mandate AA meetings. With a predictive supercomputer, it's the same thing except we do it more preemptively.

This is not wrong. Society would be much improved.

Awesome discussion.

I'm fully down with the idea of moral rules as a pragmatic way of solving a society-wide coordination game. It seems natural that such norms would have been found evolutionarily advantageous (which is one metric a game-theoretic analysis could use!).

Shaun then worries about whether this game-theoretic view means that all morality is just utilitarianism, and he worries we'll start killing one baby to spare two other babies, or whatever. But, no. The real situation is more like rule utilitarianism, where you enter into a system of constraints on your behavior because in the long run it works pretty well...

Relatedly, Will says, maybe we can use some sort of consequentialist logic (hopefully I interpreted that right) to justify mutating our norms when it appears it would be good to do so. Again, I am fully on board with this. For consequentialists (like me), "establishment of a norm" or "decision to follow a norm in this case" or "decision to punish someone" or "decision to get angry at someone" are just plain decisions, of a piece with all other decisions. All should be evaluated in a consequentialist way.

I would posit that our revulsion against utilitarian baby-killing is an outcome of some ongoing emotional training process which we are always free to do or not do. So the answer to Shaun's worry is, we don't kill a baby because it feels really rotten. From a utilitarian perspective, it is good to train ourselves to think it is rotten.

I think we have also been designed by evolution to have "feedback" from actions to moral beliefs. If I do X fifty times, X seems more OK to me now. This is critically important. This is why we do not kill babies in a utilitarian way! Because I am not just killing a baby, I am doing violence to my very system of morals. The action has longer-term feedback consequences because it will distort my moral sense. Such a distortion is (subjectively) presumptively bad. Obviously it is subjectively bad-- it is a distortion from my morals!!

Should I trust this moral inertia? Well, not always. But overall it is a very critical mechanism for ensuring moral behavior by raising the costs of deviance. I can never be deviant in a vaccuum. It changes who I am morally, and generally speaking I will never want to change who I am morally.

Is it just my imagination or dismiss the idea of fee will in this bhtv? It seems like you do - for a bit - and then decide that free will is just a fact …?

I think you just cynically summarized Daniel Dennett's entire argument in "Freedom Evolved." Ah, that's compatibilism for ya'.

Is it just my imagination or dismiss the idea of fee will in this bhtv? It seems like you do - for a bit - and then decide that free will is just a fact ...?

Is the Chomsky connection in Rawls really that peculiar...? Seems like it's well in line with Rawls's Kantianism to me.

Giving credit where credit is due: I wonder what the real story is behind Rawls's nod to Chomsky's theory of a universal syntax as a possible model for moral judgement.

The same analogy is broached by Robert Nozick in his 1967 essay "Moral Complications and Moral Structures", the only work by Nozick's referred to in all of Theory of Justice.

So who made the analogy first? Rawls or Nozick?

Never thought that I'd regret buying that book, but now I do.

You ought to get John Mikhail on. His book has been forthcoming forever - try to figure out when we'll finally get to read it!