1. We hold value X.
2. Valuing X implies that we want Y to happen.
3. Doing Z will bring about Y.
4. Therefore we will do Z.
I just made this up last week so comments/criticisms most welcome.
*****
This is a good start. But if we do Z becuase we want Y then the ends are
justifying the means. This is acceptable in a 1-D fitness landscape (for
instance):
1. I value life.
2. This implies that I want to continue to live.
3. By eating [among other things] I will continue to live.
4. Therefore I shall eat.
Simple. In such a situation there isn't really a dilemma. But want if you
only have enough food for you OR your own child? It is not so easy to derive
ought from is by this process in such a situation. What "is" constrains the
system such that all values (a plural [set] X instead of a singular X) cannot
be satisfied simultaneously. One must prioritize and compromise.
How about:
1. The universe exists in state U
2. We hold these values, X.
3. These values, X, imply that we want Y to happen.
4. Y' is the set of possible new states for which we can currently concieve of
a mechanism, Z', such that U ---> Y'.
5. Y" is a the closest approximation of Y that exists within set Y'.
6. If we do Z" then Y" will happen.
7. Therefore we will do Z".
The trouble of course, is in defining each of the terms of the equation.
Am I picking nits?
Some observations:
Is getting close to Y (ie doing Z" and getting Y") what we wanted?
What I mean is, under what cirucumstances is approximation not
acceptable?
Is U ---> Y" a step on the shortest path U --->(n steps)----> Y?
I think in this regard, line 4. of my proposal is significant: We can
cause to happen only things for which we can concieve a mechanism. It
is unlikely that in any real situation the perfect solution Z will
exist.
This is where I think about Kaufmann's idea of "drift" freeing us from local
fitness minima below the global maxima. Maybe ethics/culture/consensus
requires a certian amount of randomness in order to evolve. Complete randomness
is anarchy, but maybe a little here and there is mutation.
Also significant here is the image Kelley describes in "Out of Control" where a
room full of people fly an airplane (among other tasks) by consensus.
Every term in the above system is redefined each instant. State U is dynamic
and, in large part, unknown. Values X are provisional at best. Set Y' is
dependent on both these variables and thus fluctuates geometrically.
Have you ever played a videogame using a mouse? It's really hard. I just
can't move a mouse that quickly with precision. I keep overshooting the
target. It's a lousy control device.
But if the landscape itself is in flux, maybe a rigid-good control device isn't
what you want. Maybe you want a control device that is a little bit
in-accurate in order that you might take advantage of seredipity.
I remember Richard saying something about cost-benefit analysis in this
context. If the fitness landscape is dynamic then programming a complicated
rigid control system isn't worth it. In the extreme: if the universe is
random, then act chaotically...it's as good a bet as any.
Reed
konsler@ascat.harvard.edu