This is a letter I wrote to some friends about Mysticism.

Barring any knowledge of mathematics or psychology, why is it incorrect to postulate the existance of, e.g., a mystical "energy" to explain otherwise unexplainable observations, such as apparrent preminisions, mind-reading, etc...?

For instance, why shouldn't a child, who may have a poor grasp on statistics and no knowledge of the workings of subconscious inference, justifiably beleive in such things as E.S.P. upon personally experiencing one or two cases of apparent mind reading?

[Hmm.. no matter how I try to word it, it seems to be an obvious analog of the question "Why shouldn't I beleive in God?", but I'm going to address it from the mystic perspective as much as possible since there are an awful lot of people out there who have abandoned the god notion only to replace it with mysticism.]

First, if anybody has thoughts here, please jump in; I'm just thinking out-loud at this point. I would like some help in fleshing out this point so I can fend off mystics more efficiently in the future. :-/

Before we discard it as a trivial question, consider that a belief in mystic energies is not so different than a belief in electro-magnetic waves. Both are observable only by their effects; ultimately both must be inferred. In my experience, the present-day intelligent mystic typically feels they are on the frontier of a new science, just as the discoverers of E.M. were in times past, and that the rest of us are just close-minded conservatives. And I'm sure most of us have, as children, entertained the notion that there could be E.S.P., telekinesis, etc..., and wouldn't necessarily consider that exploratory foray to have been an artifact of bad epistemology. (This is not to say we _beleived_ in those things as children, but rather that we simply explored the possibility; perhaps therein lies the difference between the mystic and the rational.)

So, where _does_ it become bad epistemology?

Remember, again, that we are assuming little or no knowledge of statistics or psychology on the part of the mystic, so we cannot blame the error on ignoring conflicting evidence. It is hard to be fair here, because so often mystics will cite as evidence things for which we know the _rational_ explanations; but that distracts from the crux of the matter which is that the mystic's inferences are unjustified even _without_ the competing rational explanations. (I.e., surely there was a time in each of our lives when we knew as little as a mystic, and yet we didn't get our feet stuck as they have -- why not? I know a number of very intelligent people who are, nonetheless, mystic in one way or another.)

I have had a number of mystics--and even one prominent Objectivist!-- tell me, in essence, that if I could show them an alternative explanation for their observed phenomena, they might lighten their stance. This always makes me angry because they are, in essence, admitting that they beleive (enough to argue, and live by) something merely because it explains things, not because they actually understand the thing itself. (The mystics usually need a lesson in statistics and psychology; the Objectivist wanted an explanation of the introspective experience of free will.)

So, what exact step are the mystics taking that is unjustifiable? (Takers, anyone?)

As best I can tell, these mystic concepts tend to start as a kernel of related but unexplained observations, and grow into a cluster of floating abstractions attatched to the real world by this kernel. One might call the result a "hanging" abstraction. The primary fallacy is simply in ignoring Occam's razor, which would normally prune these hanging abstractions right off the tree of knowledge. But the serious consequences don't start until the direction of inference is turned around, and the hanging abstraction starts tainting expectation of actual observation. (This separation between the source of error and the consequences makes it that much harder to argue with a mystic, since the things that are obviously wrong to you and me are not necessarily irrational points within the mystic's framework.)

Specifically, I think the mystic embelishes an otherwise valid concept with desirable interpretations, and then treats these embelishments as if they were _supported_ by the observations, when in fact they are no more than arbitrary propositions (neither contradicted nor indicated by the evidence).

Let me give an example:

Let's say, in historic times, that you observe that, more often than not, a high-potency adult beverage will combust when teased with a match. In that case, it would be valid to create a concept, we'll call it "zoosh", for this combustible characteristic combined with a friendly bitter taste.

Similarly, imagine that you find, with high reliability, that you can read people's minds. That is, when you look at someone, you may feel some emotion, or hear some words in your head, or think some concept, and when you ask them about it, more often than apparent chance, they'll confirm that that was how they were feeling or what they were thinking. As with zoosh, we'll validly create a concept for our mind-reading ability, and call it "zorching".

Now, zorching and zoosh are valid concepts in the sense that you can use them to make compound observations -- e.g., don't zoosh and drive, or people who zorch make better lovers, etc...

Further, I could begin exploring hypothesis for underlying explanations which might unify these concepts with others, or give me a deeper understanding of them. E.g., I could hypothesize that the barley gets angry when you make it sit in a barrel for so long, which is why it turns bitter and becomes prone to outbursts of fire when provoked.

The actual consideration of this hypothesis is fine, so long as it serves as no more than a guide for exploration -- i.e., it might lead me to pursue beverage-grain psychology, until I have built up enough knowledge to understand the intricacies of barley-ire.

But the mystic, in historic times, might well adopt my barley-ire hypothesis as a reasonable explanation, and propigate it as gospel henceforth. Once the idea gained a following, it would be very hard to dispell, since there would be no evidence _against_ it. Further, since, as we know today, expectation leads both perception and behavior, many of the _conclusions_ drawn from the hypothesis may appear to be observably true.

The error, then, is in failing to recognize the boundary between the concept consolidating actual observations, and the hypothesis which _could lead to_ the actual observations. I.e., the justifiable concept is always constructed bottom-up, wheras the questionable hypothesis may be constructed top-down. Ergo, the hanging abstraction.

Exploratory knowledge acquisition may frequently employ hanging abstractions as guides and place-holders; but such a hypothesis should never be treated as truth until it's roots to the ground are found.

For instance, consider the aged hypothesis that electro-magnetic waves propagate through a ubiquitous substance called "ether". This is a classic example of a hanging abstraction -- it is a hypothesis of the form "if this were true, it would explain our observations." The hallmark of such an abstraction is that it adds something beyond what is directly implied by observation -- i.e., there is nothing about electromagnetic waves that necessitates an ether; ether is simply a _compatable_ hypothesis.

As an alternative to rooting such an hypothesis, the hanging abstraction can be upturned into a proper concept by observing _all_ of its direct consequences. I.e., once the complete set of observations is in place, the concept can be constructed in the normal bottom-up manner as a simple consolidation. In the case of ether, the attempt to do this disproved the hypothesis (conflicting observations were found).

The basic E.M. wave concept is an example of this up-turned hanging abstraction. E.M. waves, while their cause was, at the time, unknown, were understood as a behavioral abstraction of observation. I.e., the relatively simple set of equations which describe their behavior are, ultimately, a consolidation of observations, devoid of embelishments, with each direct implication verified by experiment. That is, the concept does not claim anything about the underlying nature of the waves, nor does it assert their behavior outside of the context in which the observations were made (e.g., at the quantum level). This is evidenced by the fact that the principles still remain valid even in light of the completely unexpected _cause_ of E.M. (i.e., quantum).

In the same sense, zoosh is a valid concept (by my original, unembelished description), so long as it goes no furthur than to consolidate observations. And, eventually, such concepts may be rooted from beneath, as we know today zoosh is alchohol, with its origins, chemical structure, behavior, etc...

To return to a more contemporary concept, consider zorching (my concept for mind-reading). Similar to zoosh, zorching is, in itself, a valid concept. The non-mystic, while he may have fanciful hypothesis about its cause, would carry the concept no further than my original description--that is, a consolidation of observations. Again, it could be applied, observed, and consolidated into other concepts; but zorching itself would remain free of embelishments.

The mystic, on the other hand, might latch onto an appealing hypothesis, and create a "cause" that may be quite consistent within the scope of the mystic's observational skills. Furthur, such a cause, if broadly enough defined, may well "explain" quite a number of phenomena (consider how much "God" explains!), which can appear to lend credence to the idea. And, again, expecation can certainly taint observation in favor of such concepts.

The difference between the hanging abstraction and a correct abstraction can be subtle; particularly in the context of the broad and vague concepts which mystics find so appealing. But the essence of it is this: A normal abstraction is created _from_ observations, and consists of _less_ than the totality of those observations (in that it is a distillation of the essentials), whereas a hanging abstraction is created _for_ the observations, and consists of _more_ than is directly implied by the observations.

Thus, the mystic can accept concepts which are so broad and vague as to defy analysis, since lacking the requirement that all legs be on the floor opens the doors for a millipede of concepts for which, by the time you find the levitating foot, you are hard-pressed to show you are on the same body.

The antithesis to the mystic is the skeptic who refuses all observations which defy explanation. I.e., he who says zorching is simply impossible! It is this side, which we may tend to fall on as a knee-jerk response to the mystics fantasies, which paints the picture of the non-mystic as being obviously closed-minded. The mystic observes it, the skeptic denies it; how simple.

The middle ground is key -- remembering that observations are axioms, and the root of _all_ explanations. Zorching may be observed, and justifiably beleived, provided it is not embelished--i.e., no explanation created for it; hence, neither a skeptic, nor a mystic be.

And from this unembelished state, the ties from other directions can be sought out objectively -- whether to pursue the possibility of mystic energies, the laws of combinatorics and chance, or the mysteries of subconscious inference.

Zorching, zoosh, and zen may all hold applicable wisdom if properly interpreted; but all can mire your feet forever if you aren't careful.

Thoughts?

-Brandyn

Date: Thu, 25 Dec 1997 13:12:31 -0800
To: hatch@hell.engr.sgi.com (Don Hatch)
From: Brandyn Webb
Subject: Re: mysticism/axioms and stuff

At 11:22 PM -0800 12/24/97, Don Hatch wrote:
>In reading your essays, my reactions seem to form a love/hate pattern
>which I thought I'd share with you for what it's worth... :-)
>
>I like, and agree with, your criticism of people who
>hold something to be true or sacred
>when its only merit is that it explains the data
>and there is no other hypothesis available.
>In particular, I like this excerpt
>from http://sifter.org/~brandyn/Mysticism.html:
>
>> In the same sense, zoosh is a valid concept (by my original, unembelished
>> description), so long as it goes no furthur than to consolidate observations.
>
>Beautiful!
>And then you go and completely blow it in the next sentence...
>
>> And, eventually, such concepts may be rooted from beneath,
>> as we know today zoosh is alchohol, with its origins,
>> chemical structure, behavior, etc... 
>
>For me, the only difference between "alcohol" and "zoosh"
>is that "alcohol" is a more intricate and sophisticated concept
>that plays well with some other concepts (like chemical structure and physical
>laws) which as a set have proved useful in further consolidating
>our observations.
>
>Fundamentally they are the same and I would apply exactly
>the same warning that you so eloquently expressed in the first
>sentence:
>	"XXX is a valid concept as long as it goes no further
>        than to consolidate observations"
>(where XXX is zoosh, alcohol, chemical and physical laws, mathematical and
>logical systems...
>*NONE* of these deserve to be considered true or sacred to my mind,
>although it's often useful to assume them in order to have a discussion).

Anyway, that's all the bathwater; back to the baby: You said it youself when you said "in further consolidating our observations". This consolidation is the "rooting" I am referring to. I detect that you are offended by the implication that there exists a rational hierarchy of concepts; but at the same time, you know (statistically) perfectly well the (observed) relationships between alcohol, its chemical properties, and zoosh. In the context of this discussion, I was contrasting against mysticism, in which _causes for_ something are invented without proper justification. My point here was to show by example a properly observed causal relationship.

Why is one a "properly observed" causal relationship and the other not? Consider:

All that can be observed is the co-occurance of things. (*) "Causality" is a statement of a particular statistical relationship, in particular:

A->B, or P(B|A) ~= 1.

(Further, in speaking of "knowledge", it is generally assumed that P(B) is less than 1, otherwise the information content of "A->B" is zero.)

Note that Baysian inference can be applied to draw (probabilistic) conclusions, such as:

[1] P(B|A)~=1, P(C|B)~=1 -> P(C|A)~=1

Note the ~ approximately equal implying always a hint of uncertainty; in particular, the above result, P(C|A)~=1, is only a statement of expectation given the observed data. Upon observing A&!C just once, the future expectation for P(C|A) may decrease significantly.

Now, consider the types of inferences from our earlier examples. To "invent" an explanation (E) for zoosh (Z), as a mystic might, is to make this inference:

E->Z -> Z->E

which is clearly incorrect. That is, just because E _would_ cause Z if E were true does not mean E must be true if Z is. There are many possible causes of Z, and the most we can say is that if E would imply Z, then Z does not imply not E -- i.e., Z and E are compatable.

Contrast this with our current scientific understanding of alcohol (A) vs. zoosh (Z). The derivation here consists _entirely_ of applications of equation [1]. That is, _observation_ shows us that one thing causes another, and these two things together cause that, and that that causes those, and so on, to the extent that it is _statistically predictable_ that certain atoms assembled into a certain configuration will have certain properties, such as burning, making people stupid, etc...

I.e., the concept of "zoosh" is thus "rooted" in lower-level observations in that it is _predictable_, or "explained", by those lower-level observations (_observations_, not inventions!). It is important to note that once zoosh itself is observed, this rooting is relatively unimportant with regards to zoosh itself, except as a new means to make predictions about zoosh which have _not_ yet been observed.

The true moral of this whole story is simply this: inference should only flow in the causal direction from observations, where causality itself is an observed statistical relationship amongst the concepts. The "we know today..." was an example of this proper chain of inference, in contrast to the hypothetical improper inference I had made on behalf of the mystics.

-Brandyn

(*) Here, co-occurance doesn't necessarily mean at the same time, but a full definition is beyond the scope of this letter; I'll flesh out the details of this in the Epistemology section of Observationalism.