[From Bill Powers (940430.0730 MDT)]
Clark McPhail (940429.2222) --
Perhaps I have missed previous discussions on this net of the
consequences of more routinely consumed drugs in this culture -
alcohol, marihuana, caffeine - for the way in which control
systems do their controlling?
There is some qualitative information about this, in the number of
alcohol-related car accidents. Unfortunately, people who haven't
been drinking also have accidents. In fact, just about any crude
qualitative measure has this drawback, in that you could probably
show a population effect of the drug, but couldn't predict whether a
given person who has taken it will lose control of a car. So the
only way to prevent alcohol-related accidents is to proscribe the
drinking of alcohol by _everybody_.
Our society seems to have gone mad lately on this approach to
understanding human functioning. We're all supposed to exercise,
avoid eating fats and other things from a long list, avoid smoke
(not to mention actually smoking), keep our cholesterol down, take
an aspirin every day, gobble beta-carotene, and so on and so on. All
this is because a statistical relationship has been found between
supposed causative factors and supposed effects in a relatively
small -- but significant -- proportion of a sample population. So to
prevent the bad effects, we are ALL supposed to follow the
prescribed rules. We even pass laws to make sure we follow some of
them.
If the right study were done, it would probably be found that there
is a significant incidence of nausea and mental upset from
ingesting, or even considering ingesting, pork products. It would
probably be found that blood pressure is elevated by putting certain
names on ballots. It would probably be found that engaging in
certain sports has an influence on the number of life-threatening
injuries that occur, both among participants and spectators. With
large enough studies, these effects would become incontrovertible.
What then? Laws against pork products, certain names on ballots, and
certain sports like soccer and car racing?
In a series of studies done in the 1970s a psychologist at Kent
State University examined the retaliation for electric shocks
received by subjects who had consumed either alcohol or
marihuana. Alcohol consumers retaliated more quickly and by
administering more shock in return than did marihuana
consumers. For alcohol consumers the relationship was
curvilinear with retaliatory shock increasing with the increase
in alcohol consumed but only up to a point after which time the
alcohol consumers retaliated less and less.
I echo Tom Bourbon's question, in slightly different words. If you
picked a person at random from the group studied, could you predict
whether this person would retaliate more or less when given alcohol
or marijuana than when not given it? Phil Runkel has had a lot to
say about this sort of study -- a study in which people are assumed
to be interchangeable, so that one person on alcohol can be compared
with a different person on marijuana and a third on nothing, and in
which a group trend is assumed to represent individual
characteristics.
Since retaliation can be seen as a form of disturbance-resistance
(encouraging the shock-giver to stop giving shocks), how are we to
interpret the facts, even accepting that they are facts? If a person
on alcohol retaliates more for being shocked than someone else under
different conditions does, does this mean that the person on alcohol
is controlling worse, or better than the other person? Could we say
that a sober person who does not retaliate is more vulnerable to
being shocked? Maybe, in the interests of self-preservation, we all
should walk around with a mild load on.
The approach I would like to see is to measure individual
performance as a function of drug ingestion, and to measure it in
terms of changes in the best-fit parameters of a model of the
behaving system. This depends, of course, on HAVING a workable model
of the behaving system. Unfortunately, research money does not go
very generously into building models. Building workable models is a
long slow process with little interesting payoff in the early
stages.
Drug companies can't make a profit by studying WHY drugs have the
effects they have on some people, and why they have different
effects or no effects on others. In fact, big statistical studies
are beneficial to a drug company. The bigger the study, the less
effect a drug has to have to demonstrate a significant benefit. So
by doing large impressive studies, you can convince large numbers of
people that they should buy and use the drug even when it is far
more likely to have no effect or only a "side-effect" than to help a
given person. It's ironic, isn't it? The larger the population
studied, the less likely it is that the effect found is important to
any given person. You do large studies only when the effect is too
small to be found significant in a smaller study.
Incidentally, Thomas J. Moore has published a new book: _Lifespan:
who lives longer and why_. Among other gems, he found a list showing
the contributions to number of deaths by heart attacks from a list
of risk factors as determined by the NIH. The total came to about a
million deaths per year attributable to these risk factors. That was
twice the number of actual deaths due to heart attacks, and about 20
times the number of deaths _indepedently_ attributable to those
factors. Statistical madness.
The same Tom Moore, in an earlier book called _Heart Failure_,
showed that when the cholesterol flap occurred, the researchers
arbitrarily decided that the upper quarter of the people showing the
highest cholesterol were at immediate risk of heart-attacks and
other diseases. So at one stroke, they declared that a quarter of
the population of the United States had a life-threatening condition
requiring immediate treatment by a physician. What a gold-mine,
while it lasted.
Our Thought Policeman is temporarily in self-imposed exile. But
somebody has to do the job.
ยทยทยท
--------------------------------------------------------------
Best,
Bill P.