[From Bruce Gregory (980508.0925 EDT)]
If we do in fact always act to reduce overall system error, we can
understand why punishments and rewards are often ineffective. Punishments
and rewards are efforts to alter actions by tinkering with a subset of the
overall hierarchy. They will _only_ work if this tinkering reduces total
system error. In light of the complexity of the total system, it is very
difficult to predict the effects of such interventions.
Best Offer