Linking; statistics

[From Bill Powers (920925.1100)]

Martin Taylor (920925.1000) --

My objection to invented terms like "linking" is that it leads too
easily to logical errors, particularly in verbal discourse (as opposed
to mathematical). What happens is that one phenomenon is given the
name linking; call it linking-A. Then another phenomenon is called
linking-B. Aha, says the brain, they're both "linking." Therefore the
first phenomenon must be just like the second. I gave Greg an example
of a different kind of linking that was not at all like the kind he
was thinking of.

This is what I've been calling the "category error." It's my main
objection to verbal generalizing as a mode of theorizing. By
categorizing, we strip away differences, so we can say that phenomena
that are actually quite different "can be seen as" examples of the
same thing -- that is, members of the same category. But this works
only one way; you can't say that because A belongs to category C and B
belongs to category C that A and B can be treated interchangeably from
then on. Particularly when only words are involved, this is really a
form of punning. My cat is attached to mice and my computer is
attached to mice. Therefore cats are like computers.

This error is committed ALL THE TIME in psychology. It's compounded by
saying that people do certain similar things that can be seen as
category A, under various circumstances that can all be seen as
belonging to category B; therefore, circumstances in category B result
in doing things in category A. Now we have categories causing
categories. This is one of the main reasons that theories in
psychology predict so poorly; this sort of category error generates
huge numbers of counterexamples when you try to reduce the
generalization to specific cases again. As long as you STAY at the
category level, the generalizations continue to hold more or less
true. But any attempt to apply them in real cases will succeed only by
luck. That's because generalizations have no causal force in nature;
they're only conclusions by an observer. By the way, they do a lot of
this in physics, too, although mathematics helps to limit the damage.

···

-------------------------------------------------

When I said Pat couldn't control for her kids' health because her
knowledge of her effect on their health was statistical, I didn't mean
that her perception of the effects was statistical. Generally, there
isn't any perception of the effects at all. You can't see the effect
of feeding good food to the kids unless you sometimes feed them bad
food, too, and observe the difference. The statistics I was referring
to was the "findings" of nutritional science, which deal with
populations, not Pat's kids. We (to take the onus off Pat) feed our
kids food that we understand to be good for them, for personal
theoretical reasons and because of what other people have found
through nutritional studies of populations. This has to be mostly an
open-loop process. We don't know if our own kids are typical, and we
don't have the means of monitoring the effects that are available to a
research laboratory. Mainly, though, we don't have control over their
health because we don't experiment with it; we don't really know
whether _varying_ their nutrition would _vary_ their health. If the
kids look poorly, we'll look for other causes, not at what we're
feeding them (unless we decide that it wasn't nutritional after all --
but then when they get well we'll stick to the new diet, not go back
to the old one). Pat's not going to put her kids on a diet of Big Macs
and fries for a month to vindicate her views on nutrition, and if they
get sick and recover, she's not going to decide that their former diet
made them sick (healthy food doesn't make people sick).

There are lots of things we have to do open loop (or think we have to)
because we really don't have any direct control over them. This is
what makes medicine, psychotherapy, playing the stock market, and
cleaning the chimney every fall such grey areas -- did we have an
effect, or was it something else? Or would the same thing have
happened no matter what we did? We can't just let the world happen
without trying to do SOMETHING about it, can we? So we go through the
motions anyway, hoping to have an effect even if we'll never know what
it was.
------------------------------------------------------------------
Ray Allis (920925.0900)--
RE: turing test

If you could not distinguish whether the entity, with which you
would communicate by typewriter, was human or not then it must be
able to think. Sort of an early version of "If it walks like a
duck, and quacks like a duck..."

As you can see, this obviates the need to say what is meant by
"think".

Right on. All the Turing test can really ask is whether the other
entity is a human being or a program (or, I suppose, a duck).
------------------------------------------------------------------
Best,

Bill P.

[Martin Taylor 920925 14:30 --finally all caught up!!!]
(Bill Powers 920925.1100)

We (to take the onus off Pat) feed our
kids food that we understand to be good for them, for personal
theoretical reasons and because of what other people have found
through nutritional studies of populations. This has to be mostly an
open-loop process. We don't know if our own kids are typical, and we
don't have the means of monitoring the effects that are available to a
research laboratory. Mainly, though, we don't have control over their
health because we don't experiment with it; we don't really know
whether _varying_ their nutrition would _vary_ their health. If the
kids look poorly, we'll look for other causes, not at what we're
feeding them (unless we decide that it wasn't nutritional after all --
but then when they get well we'll stick to the new diet, not go back
to the old one).

One doesn't reorganize when there isn't error. Sure. But for a lot of
parents with unhealthy children, changes in diet often are among the changes
that they try, until they find what the kid was (say) allergic to. The kid
becomes healthier, and they no longer try to change the diet. The effect
may have been what Skinner called "superstition"--something else caused the
good result by coincidence, but nevertheless, the error has gone away, and
reorganization stops. No doubt diet change was not the only action involved
in the reorganization process, but it was the perception of the kid's health
that was the subject of control, and the perception of the relation between
the parent's action and the child's health is itself statistical. There is,
to help, the population-statistical information about the probabilities that
people with certain diets more often have certain symptoms than do people
with other diets. That information can help the parent to choose which diet
changes to try, and to that degree, perhaps the reorganization is not
random. Maybe this relates to Greg's "guided reorganization?" We are getting
into a murky area here, I think.

···

----------

My objection to invented terms like "linking" is that it leads too
easily to logical errors, particularly in verbal discourse (as opposed
to mathematical). What happens is that one phenomenon is given the
name linking; call it linking-A. Then another phenomenon is called
linking-B. Aha, says the brain, they're both "linking." Therefore the
first phenomenon must be just like the second. I gave Greg an example
of a different kind of linking that was not at all like the kind he
was thinking of.

"I have to use words when I talk to you" at least by e-mail. I don't think
our mathematical descriptions of the feedback loops are either accurate or
would be intelligible if they were complete. This is especially true since
the feedback loop incorporates the disturbed world.

We use "reference" and "percept" and "error" with technical meaning. Why not
"linking" which seems to describe a phenomenon that is important in
interpersonal interaction. I'm not sure it occurs when only one hierarchic
control system interacts with the inanimate world. It's an extra phenomenon
and need a word. "Linking" seems to fill the bill.

I agree that verbal puns can lead one's theorizing astray. It happens a lot.
But often the verbal mode is the best we can do. I doubt you could mathematize
Ed Ford's juvenile delinquents very effectively.

Say "OWA TANA SIAM" and see the six-year-old laugh.

Martin