[From Rick Marken (960206.2100)]
Lars Christian Smith (5 Feb 96 08:10 CET) --
Learning by doing (originally described by Kenneth Arrow in his 1962
paper, 'The Economic Implications of Learning by Doing') describes how the
unit cost of production is a function of total cumulative output. As
individuals gain experience in undertaking some repetetive task, their
efficency in performing this task increases and the time taken to
complete the task decreases.
How would you think about this phenomenon in HPCT terms? Would you get
any additional insights by thinking about it in HPCT terms, e.g. in
terms of reorganization? How would you model it?
I'd think about this in terms of what I know about reorganization and
hierarchical control. First, I wouldn't expect the increase in the
efficiency of task performance to _necessarily_ be regular;I would
expect "plateaus" of indeterminate duration where the "solution" that
has been hit upon by the reorganization system is "good enough". I'd
also expect _bouts_ of reorganization during which task performance
deteriorates; these bouts of reorganization may be followed by improved
performance (if the reorganization system hits on a new and better way
to produce the desired results) or by the same level of performance
as what occurred prior to the period of reorganization.
I would also expect the reorganization system to settle on a means
of controlling that is best from the actor's perspective -- but not
necessarily the one that is the best possible from an objective
(observer's) perspective. An observer may know that it is possible to
assemble transmissions in 5 minutes but not everybody may end up
reorganizing their transmission assembly control systems to achieve
results this efficiently.
Finally, I would expect that the person doing the task is controlling
_many other_ variables at the same time. At some point, I would expect
that a person's reference for producing the results that we see as
"performing the task" would have to change (even if the person is
well-fed and physically able to perform the task); the person would
be seen as no longer wanting to do the task -- for the time being.
It would be hard to model reorganization unless the variable to be
controlled in the task were very somple and clear. Modelling
reorganization -- real reorganization -- requires (at least) that we
know what the system might end up controlling when it is reorganized.
Reorganization can be modelled -- I think Tom Bourbon built the most
realistic model (in terms of matching human behavior) and Bill Powers
built the most complex one -- but it's not a walk in the park.
Best
Rick