REPLY Why don't organizations function better? (SD6614)

[SDMAIL Jim Hines]

Posted by  "Jim Hines" <jim at ventanasystems.com>

Wade, 

1.	Too much mutation prevents evolution of complex structures (this is
what I said in my original posting).  That point is just not controversial -
a mutation in a complex structure is more likely to be harmful than
beneficial.  That's why prospective parents worry about babies.  

2.	While too much mutation is bad; a little tiny bit is necessary for
evolution.  Mutation creates the diversity that selection needs to operate.

3.	Recombination is very important.  It speeds up evolution enormously.
(Low rates of) mutation plus selection produces slow evolution.  (Low rates
of) mutation + recombination + selection produces much faster evolution.
(Biological evolution with recombination is still slow relative to most
human time-spans, but without recombination evolution would be MUCH smaller.

4.	The above points are easily demonstrable in a genetic algorithm.
Holland knows all of this and has written about it.  

The analogy between biological and organizational evolution is a deep one,
but it's necessary to be precise about the analogy.  Loose words like
"change" don't help.  Here's the analogy for evolution within an
organization:

Gene = Policy (in the SD sense - i.e. decision rule)

Recombination = Imitation (of a policy)

Mutation = Inexact imitation or intentional change (for any reason including
intended improvement - the idea is that organizations are so complicated
that any change has intended improvement has a significant "random"
component).

Selection = Organizational "pointing and pushing devices" (e.g. promotion or
issuing gold stars or whatever).

For more, once again, see http://web.mit.edu/org-ev/www/ and click on the
"publications" link.

Your opinion that organizations do change slowly is quite legitimate.  I
know what you mean.  However, consider what the biological example would
mean in an organization.  If the average  gene is 10 thousand base pairs
long, a gene can be replicated about 10,000 time with only ONE change.  Do
you think a manager could communicate, say, his pricing policy to 10,000
other managers so that 9,999 of them could duplicate it exactly?

Jim
Posted by  "Jim Hines" <jim at ventanasystems.com>
posting date  Tue, 11 Sep 2007 14:01:48 -0500