[From Bill Powers (950104.0845 MST)]
Rick Marken (950103.1530 PST) --
Jeff Vancouver (950103.1530) --
Jeff's huge paper is very impressive, but see my earlier post this
morning. Bringing together dozens of different approaches to living
systems assumes that all these approaches are compatible: that the
result of combining them will yield a net gain. I think that the
synthesis achievable in this way is mostly cosmetic: it makes these
approaches seem much more coordinated than they really are.
As to decision-making, it struck me yesterday that the reason people
work out "decision making methods" is in order NOT to have to make any
decision. If you do a cost-benefit analysis on method A and method B,
you come up with a C/B ratio for each one and pick the method with the
lower ratio. Where's the decision? The only time you really have to make
a decision is when the numbers come out exactly equal.
If choice of a course of action depends on a number of factors, and if
there is some rationale, reason, or algorithm involve in weighing the
factors, then the course of action that will occur is foreordained in
the factors taken into account and the reasoning applied to them. There
is neither a choice nor a decision required: just turn the crank and out
comes the answer. You don't need any machinery that isn't in HPCT.
True decisions or choices, as Mary pointed out some time ago, are
_conflicts_. They occur only when higher-order systems do not
automatically select one action rather than another, but try to achieve
incompatible goals. Now there is no systematic way to resolve the
conflict: if you say eenie-meenie-minie-mo and end up on either side,
the other side objects and nothing happens. The only solution is to
reorganize the systems that are causing the conflict, and that involves
random variations, not systematic methods.
Decision _making_ methods are really decision _avoiding_ methods.
···
---------------------------
RE: demo model
I think virtually all conventional research is motivated by the
observation (or the hope) that some variable in the environment
affects some behavior; the obvious effects of behavior on the
environment typically go unnoticed.
If the demo were presented to a conventional behavioral scientist, all
that would be visible would be the visual phenomena on the screen and
the movements of the mouse. There might be some vague realization that
the mouse is affecting the display, but that is true of practically any
behavioral situation and I don't think it would have much effect. What I
would predict is that the display variables would be taken as
independent variables and the mouse behavior as the dependent variable.
Even an a-causal ANOVA would involve calculating correlations between
stimuli and responses, wouldn't it? How likely do you think it is that
any conventional scientist analyzing this situation would realize that
there is an independent but invisible disturbance acting which
correlates almost perfectly with the mouse movements? I think that the
conventional approach would end up attributing the mouse movements to
the changes in the two objects that are NOT under control; they would
contribute by far the most to "explaining the variance" of the mouse
positions.
Of course this could also be done just by using three cursors and three
stationary targets, but I thought that having different kinds of control
tasks would get around the objection that this is "just tracking."
We'll see what Bruce Abbott has to say.
---------------------------------------------------------------------
Best to all,
Bill P.