This is part of a letter I wrote to some friends during a discussion of Free Will, along with a more recent reply from David Noelle

The point here, as relates to Free Will (and many other debates) is not just that semantics are imprecise, but that the errors in semantics, and other errors of logic, can bleed back into our "intuitive" thinking and continue to cause us problems even when we think we've set them aside. For instance, I can say, accurrately, that I am free to choose what I want to choose. And then I can engage my semantic-logical mind and say if I am free to choose than I could have chosen something else, so the universe cannot possibly be predetermined. How can you argue that?

But all of these concepts I'm working with, just like "Life", are not as simple and context-free as we are assuming them to be. Context is key -- do we really know what we mean when we say "pre-determined", "free", "choice", "want", and, worst of all, "I"? I think with a little consideration, you'll see that the two clauses simply do not exist within the same context. Determinism exists at a sub-chaotic level, where every molecule is considerred and is relevant. "I" exists at a super-chaotic level, and _requires_ ignorance of the underlying mechanics. "I" is, by my definition anyway, the information process above the mechanics. In Objectivist terminology, many measurements have been ommitted.

So how do people so easily bridge the two contexts? I think it usually happens at the concept "pre-determined". Pre-determination is semantically two-faced, in that it has the same meaning at both levels (both contexts), and yet the two uses are nonetheless separable. To be specific: to say an object's time-course is pre-determined is to say that _within the constraints of measurements possible within that object's context_ the future state of that object is theoretically predictable. Let me give my favorite example of this same law operating in two contexts on the same object and producing the opposite answer: Imagine a clock-gear which is composed at the molecular level of stochastically wiggling molecules. Speaking of the clock gear, we can say accurately that it's course is pre-determined -- we know how far it will rotate in any given amount of time. But speaking of all of the molecules that makes up the clock gear (it's the same thing, right?) we can accurately say it is NOT pre-determined, because we have no way of knowing where exactly each molecule will be in it's random, wiggly path. To say the clock gear was therefor no deterministic would be an error! Likewise to the converse.

The free-will case is the juxtaposition of the clock gear. That is, if we adopt the hypothesis that the universe _is_ deterministic for the sake of argument, we have a case of a non-deterministic process built upon a deterministic one. At the level of molecules, everything is set in stone. But at the level of consciousness, or even of the macroscopic world in general, the universe is hardly deterministic! I can choose what I want to do -- there is no arguing that. My life is not set in stone -- I could decide tomorrow to go do something completely different. It _is_ up to _me_ -- within the constraints of the measurements possible within my context. I.e., according to what I can observe; according to what I can introspect. ** In order to make the measurement that would demonstrate the deterministic nature of the universe, I would have to shift to a context in which the concept of "I" no longer exists (remember the clock gear). ** So the universe _could_ be at once microscopically determined and macroscopically non-determined, and hence I can have "free will" with all of it's ramifications while still knowing that at the microscopic level my fate may be sealed.

That last sentence has a big clue: "with all of it's ramifications." Think of each and every practical application of the concept of free will, and tell me if you find one that applies at all at the microscopic level. Keep in mind that ethics is entirely defined within the high-level context of consciousness and macroscopic behavior.

So the last question remains: having seen how the determinism clause spans the two contexts, how does the free-will side work? All of the concepts in question, "I", "choose", and "want" rest on the boundary between the two contexts. The introspective versions lie on the top side-- I am an information process interfacing with the world at a very macroscopic, macro-conceptual level. My wants are felt, my choices are made, and I am. But for each of these, there must in turn be a cause, and thus each has an underside -- the dual context as with the clock gear -- which implements the abstraction. My wants are computed, my choices are executed, and I am the sum of those and other things. It can't be turtles all the way down. Which leads to the final question:

I will agree you are free to choose what you _want to choose_. But are you free to choose what you _want_ to choose? And if so, are you free to choose what to want to choose? And so on? (Turtles all the way down?)

Date: Wed, 7 Feb 96 19:44:28 PST
From: brandyn (Brandyn)
Subject: more Free Will (trying to clarify my last letter)

[This is a little long, but I tried to wrap it up tightly at the end..]

[...] Someone complained of my apparent implication that: Since a deterministic thing (the gear) could be constructed from non-deterministic components (the quantum particles, or whatever), the converse must also be true (the non-deterministic process of free-will could be created from the deterministic constituents [neurons, whatever]). I didn't mean to make this implication -- one does not directly imply the other. Although I still do consider both to be true, but with independent proofs. Is the gear example clear?

Dice are a trivial example of non-deterministic objects built on (potentially) deterministic components. The weather is a less trivial example.

But is this "non-deterministic" nature of the weather just an epistemological construct, or does it have any real-world basis? I say determinism, as an attribute of an object or system, is very much a context-sensitive attribute, with real-world consequences, and so the "deterministic character" of a thing is truely level-dependent. I.e., you could say that dice are deterministic at the molecular level, but non-deterministic at the _dice_ level. And this is _not_ just an epistemological construct -- anything that was dealing with those object as _dice_ would almost certainly receive them _as non-deterministic objects_. I.e., if I gave you a pair of dice that were truely non-deterministic all the way down (such that if the quantum context were IDENTICAL on two throws, they still may turn up differently) would you know the difference? Would a machine or process that depended on their behavior behave any differently? No! When objects combine to form a compound object, the properties of that compound object are determined by the _properties_ of the constituent objects, and then only _certain properties_ of those objects, and those certain properties nearly always exist within a limitted context or level of abstraction.

I'm having a hard time not appealing to my structure-in-the-universe argument, which is an entirely different matter, but adds much light to this determinism question... The punch line of that argument is that our concepts, rather than being somewhat arbitrarily demarked as many people beleive, are in fact mirrorring _inherent_ structure in the universe, defined by the non-linear nature of emergent properties accross levels of abstraction. Ergo, principles which exist at multiple levels of abstraction, such a determinism, addition, spacial transformations, etc.. _apply only within one level of abstraction at a time_. For instance, you can rotate an object without being concerned whether all of its molecules rotate too, or whether some of them just translate while maintaining their same rotation. Thus, the orientation of the higher-level object is a property unique from the orientations of its constituent objects, even though it is in fact the same principle applying in both cases. Determinism works like this. To say something is deterministic or non-deterministic within a particular context is _not_ to say it or its constituents or amalgams are so in all contexts.

Now... after all that... back to the point: by any proper (by my claim) and useful definition of determinism, consciousness is non-deterministic at _our_ level of abstraction. Consider Rick's note that it is determined by experience, learning, background, etc.., I don't dissagree with this _at the lowest levels_, but even taking all external factors into account, one still could not predict which way the pendulum might fall on any particular matter without knowing all of the internal state of the machine, exactly (that's Chaos for you). The introspection of free will is an indicator of that -- we all teeter on decisions from time to time, and I don't think even Rick would claim that the final throw is always _determined_ by external or historical influences, but might rest on something so arbitrary as the phase of your blood pulse with respect to your alpha rhythms or whatever [sorry if I'm slaughtering terms here].

Am I making any sense with any of this, by the way, or do you all think I'm a lunatic?

Anyway, back one more level -- if you recall, my original claim was that free will is not incompatable with hard-determinism. So to clarify what I mean by that in light of my newly expounded definitions: Free will, as is implied by its name, implications, and introspection, requires non-determinism _at the level of abstraction in which consciousness is defined and meaningful_. Furthur, this non-determinism is _not_ incompatable with molecular-level determinism, _as I have defined determinism, and as the characteristics of determinism are exhibited_.

To wrap it up: Even if we knew the universe were molecularly deterministic, we could predict from the laws of physics and chaos that there would be non-deterministic processes at higher levels of abstraction, and so it should be no surprise that we find an example of this in our own consciousness. Ergo, free will does not _presuppose_ molecular non-determinism.

Objections? Seem reasonable?

-Brandyn

Date: Sat, 11 Apr 1998 00:19:42 -0400 (EDT)
From: David Noelle
Subject: Free Will

Brandyn,

[...]

I am rather fond of the "different levels of analysis" position in this matter. Rather than listing all of the other things about your essay with which I concur, however, I will immediately address the main problems that I had with your argument.

First, randomness is not enough to make for what most people consider to be "free will". Thus, the potential presence of non-determinism in the form of a random causal factor is not sufficient to explain freedom of action. Am I any more free if my actions are caused by a dice throw than if they are caused by a deterministic process? In either case, traditional notions of free choice do not enter into the picture.

Second, even if you believe that randomness is enough to account for "free will", it is not clear to me how such true randomness could arise from a perfectly deterministic process. (Now, I have a loose understanding of non-determinism at the subatomic level, and I know that this source of randomness might be drawn upon to explain randomness at a macroscopic scale. Still, your essay seemed to argue that true randomness might arise from a purely deterministic world.) In the case of your clock gear composed of stochastically moving particles, the reduction from a random process to an essentially non-random process is clear. If we were to examine the joint probability distribution of the locations of all of the component particles, there would be a big spike around the joint locations which give rise to a gear shape. If such a spike was not present in the probability distribution, the gear would quickly cease to be a gear. In short, we have a rather precise way of relating the different levels of analysis in this case. When trying to construct randomness from a deterministic world, however, I am at a loss as to how such a reduction might be done. Some appeal to the chaos of complexity theory seems like a good place to look, but I don't think it provides an answer. A deterministic function can certainly exhibit pseudo-randomness but never true randomness. Indeed, to identify the macroscopic result of, say, a dice throw as any more random than the underlying (assumed) deterministic world is merely a statement of one's ignorance. I am ignorant of all of the causal factors playing on the dice throw, and so the result is not predictable *to me*, and so I will call the result "random". Even the unpredictability of so-called chaotic functions is really a statement of ignorance -- ignorance of the *exact* values of the state variables. In other words, invoking complexity theory is just to explain how easy it is for (deterministic) causes to be hidden from our view.

I guess I still think that "free will" is a fancy term for complex cogitations that we experience through introspection prior to action. I suspect that these cogitations are essentially determined by the structure of our brain and its current environment, but, because of our ignorance, we are unable to predict the outcome of these cogitations prior to their completion. In other words, "free will" is a label for deterministic causes of our actions which are hidden from our view.

I hope that this helps or is at least entertaining, and I hope you are well.

----------------------------------------------------------------
-- David Noelle ---- Center for the Neural Basis of Cognition --
-- noelle@acm.org -- http://www.cnbc.cmu.edu/~noelle/ ----------
----------------------------------------------------------------

Date: Sat, 11 Apr 1998 00:05:51 -0700
From: Brandyn Webb 
To: David Noelle 
Subject: Re: Free Will

>I guess I still think that "free will" is a fancy term for complex
>cogitations that we experience through introspection prior to action.
>I suspect that these cogitations are essentially determined by the
>structure of our brain and its current environment, but, because of
>our ignorance, we are unable to predict the outcome of these
>cogitations prior to their completion.  In other words, "free will" is
>a label for deterministic causes of our actions which are hidden from
>our view.

I agree completely. My point about the levels was to say, in essence, that our necessary ignorance of the micro mechanics of our own brains makes the determinism or non-determinism of those mechanics irrelevant at our level of existence. And that, _in effect_, our wills and lives are non-deterministic in the same sense as a pair of dice. I.e., I agree completely that if we are molecularly determined, then given perfect information our futures are completely mapped out (that is a tautology). However, even in that case, I would argue that there is a meaningful application of the concept and term of non-determinism, as applies _at our level of ignorance_.

I.e., determinism is a concept or measure that can be applied to any system of variables. And the measure it gives is a measure _of that set of variables_. The question is, simply: Given the exact state of these variables, can you predict their state arbitrarily far into the future? If yes, then the system is deterministic. If no, then it is not. (And I suppose you could have grey zones inbetween as well.) If you choose a different subset of variables, you may get a very different answer. And these are the levels I'm talking about: which variables are you looking at?

This "softer" use of the word "determinism" is, I would argue, the only consistent interpretation. The alternative is to assume that you are somehow talking about _all_ the variables. I.e., to say the universe is "deterministic" in the strictest sense would be to claim that, for instance, there was nothing else going on inside the quarks. What if they're made of randomly jiggling things which don't happen to effect their overall behavior? You could say "well, if it doesn't effect their behavior, then it's irrelevant" and that's just my point: you choose the set of variables you are concerned with (and specifically we tend to try to choose the set of variables that is deterministic because the other ones are a pain in the ass). But if those jiggly quark innards were truly random, then we'd have to say the _universe_ was non-deterministic; but this would obfuscate the deterministic (my definition) nature of the quarks. Follow? (Of course, I've chosen a bad example because quarks don't behave deterministically; but I'm speaking hypothetically here as if they did..)

So, here's the real value of this distinction: I see people implcitly applying "determinism" to very different sets of variables, without explicitly acknowledging that they're talking about two completely different things. Most people who argue _for_ determinism are speaking of _molecular_ determinism, whereas most people who argue against it are speaking of _perceptual_ determinism. That is, the latter group is concerned with all of the variables that are directly and practically accessible to us, not the ones that a quantum supercomputing neuroscientist might think about.

In my experience, the pro non-determinism camp is entirely motivated by two things: the need for ethics, and the emotional need for autonomy. If you tell them the universe is pre-determined, this violates both of those needs -- how can you blame anyone for what they had no choice but to do, and how come I seem to be able to do whatever I want? My answer to them is simply this: If you take every variable into account that you will ever personally encounter, that system of variables--which defines the world _we_, human consciousnesses, live in--is a non-deterministic system. And, perhaps more to the point, _those_ are the variables by which you measure ethics and personal autonomy--by what someone does, macroscopically, not by the fact that their hydrogen atoms act just like anyone else's.

So, in sum: At the level at which we exist (selecting for the variables which are accessible to us), the universe is non-deterministic. And, by the definition I have given it, that does not preclude (nor necessitate) the same universe from being molecularly determined.

-Brandyn

Date: Sat, 11 Apr 1998 18:39:39 -0400 (EDT)
From: David Noelle 
To: brandyn@sifter.org
Subject: Re: Free Will

Brandyn,

> 	In my experience, the pro non-determinism camp is entirely
> motivated by two things: the need for ethics, and the emotional need
> for autonomy.  If you tell them the universe is pre-determined, this
> violates both of those needs -- how can you blame anyone for what they
> had no choice but to do, and how come I seem to be able to do whatever
> I want?  My answer to them is simply this:  If you take every variable
> into account that you will ever personally encounter, that system of
> variables--which defines the world _we_, human consciousnesses,
> live in--is a non-deterministic system.  And, perhaps more to the point,
> _those_ are the variables by which you measure ethics and personal
> autonomy--by what someone does, macroscopically, not by the fact that
> their hydrogen atoms act just like anyone else's.

Yes. This is the real center of this debate. I would phrase it as a need to justify praise and blame. Unfortunately, identifying human behavior as non-deterministic (at the relevant level of analysis) doesn't seem to be enough to satisfy these needs. Am I justified in assigning praise and blame for actions which are the result of a random, rather than determined, process? This is an echo of the first point from my last message. Non-determinism at the level of personal action is not enough. Indeed, the whole determinism versus non-determinism debate may be irrelevant to these needs.

Instead, I think we have to get comfortable with the idea that it is acceptable to praise and blame "bottlenecks" in a causal chain. Let's say that my copy machine goes on the fritz and starts pumping out only completely black pages. What caused this? It might partially be the responsibility of the copy machine design team, who did not make the machine robust enough. Perhaps some blame may be placed on the person who bought the machine, for not selecting a better model. Maybe the office staff in charge of maintenance could be seen as partially at fault. Regardless of the details, one thing is certain: The copy machine is broken. I can point at this machine and say, "Bad!" In short, I can "blame" the machine in the sense that correcting the situation is best handled in the immediate term by applying remedial operations to the machine itself. Similarly, even if the behavior of people is perfectly deterministic, we can still place blame on them, in the same sense. We can see the person as "broken", as needing "repairs" and/or "replacement". The best way to remedy the situation is for the person to change. This sounds cold, but it is the only way that I have come across which faithfully reconciles the need for praise/blame and the way in which human action seems to arise from a mix of deterministic and random processes.

[...]

----------------------------------------------------------------
-- David Noelle ---- Center for the Neural Basis of Cognition --
-- noelle@acm.org -- http://www.cnbc.cmu.edu/~noelle/ ----------
----------------------------------------------------------------

Date: Sat, 11 Apr 1998 17:20:40 -0700
From: Brandyn Webb 
To: David Noelle 
Subject: Re: Free Will

>[...] Unfortunately, identifying human
>behavior as non-deterministic (at the relevant level of analysis)
>doesn't seem to be enough to satisfy these needs.  Am I justified in
>assigning praise and blame for actions which are the result of a
>random, rather than determined, process?  This is an echo of the first
>point from my last message.  Non-determinism at the level of personal
>action is not enough.  Indeed, the whole determinism versus
>non-determinism debate may be irrelevant to these needs.

Yes, agreed, in the sense you mean it. But, once again: revisit the meaning of the term non-deterministic. You keep referring to it as random, whereas I simply think of it as autonomous from the system. I.e., by selecting our set of variables, we declare an ignorance that makes the selected variables non-deterministic -- that is, we cannot predict their future behavior simply from what we know about them now. But this does not mean the mechanisms behind the variables are random -- it just means the mechanism is hidden. So in this way we can sort of have our cake and eat it too -- we humans are non-determined, at the observable level, because we are comprised of mechanisms which are autonomous from the observable set of variables.

We are saying the same thing. I'm just trying to pound the definition of determinism into a more useful shape.

>Instead, I think we have to get comfortable with the idea that it is
>acceptable to praise and blame "bottlenecks" in a causal chain.

Once again, agreed. Or perhaps I'll propose a slightly different angle: praising and blaming mechanisms. Really, that is what it is all about, isn't it? So say we're made of little gears and pullies that just follow some predetermined set of hard-and-fast rules. So be it, but then _that_ is _us_ -- i.e., I _am_ that system of gears (or, more accurately, my consciousness is manifest in the behavior of that system). So if I go around killing people, blame it on the organization of my gears and pullies -- but that means blame it on _me_!

>Similarly, even if the behavior of
>people is perfectly deterministic, we can still place blame on them,
>in the same sense.  We can see the person as "broken", as needing
>"repairs" and/or "replacement".

Exactly! I guess we are saying the same thing after all.

:)

-Brandyn