From: John Conover <john@email.johncon.com>

Subject: Re: When are SD models appropriate?

Date: Sat, 17 Dec 94 23:58 PST

Stephen Robbins writes: > > Having just completed MIT's course in system dynamics, my impression > is that system dynamic models are appropriate whenever you have > multiple feedback loops in a system, and/or long delays between cause > and effect. What the models helped me understand most in the course > of the semester was the overall behavior patterns of systems. For > example, we built a model of a particular marketplace, and discovered > that, given the model's assumption, there were certain policies a > company could follow to guarantee a profit in that marketplace under > a wide variety of circumstances. [It also showed us that the market > had a prisoner's dilemma characteristic--we could win BIG if we ^^^^^^^^^^^^^^^^^^ Hi Stever. What is a prisoner's dilemma characteristic in the market place? How do you run a defection strategy in the market place. Don't all competitors in a market place run a defection strategy? Isn't that the essence of capitalism? Isn't a cooperation strategy illegal, and a violation of SEC rules and regulations? (BTW, it is not clear to me that markets are zero sum, although a lot of folks pay lip service to the buz words.) Note that the classical prisoner's dilemma (as per Luce, et al) is a SINGLE iterated "game," between only two players, which probably does not model the market dynamics adequately. When there are multiple iterated "games" available concurrently in the marketplace, the situation is much different (as per Axelrod and Forrest,) since you can choose who your opponent is, and perhaps find an opponent that will run a cooperation strategy, eg, form a coalition (at least for a while.) Thus, emergent phenomena are an expected outcome of the market dynamics. IMHO, that emergent phenomena is how markets develop. Wana good PHD thesis? Most markets, empirically, exhibit a 1/f squared time series power spectrum distribution (sometimes cubed, etc.) Prove that a market with an alternative of multiple iterated prisoner's dilemma games will exhibit a 1/f type of time series power spectrum. (If, indeed, a prisoner's dilemma type of scenario is the "engine" of a competitive market place-which may be counter intuitive to classical and neo-classical economics and its equilibrium paradigm.) How does all of this relate to learning organizations? If you get the program fractal dimension program, FD3 (available via anonymous ftp from lyapunov.ucsd.edu:/pub/cal-state-stan [132.239.86.10],) and run it on the monthly statistics for the electronic components market (I work in the semiconductor industry, and it is available on electronic media from the Department of Commerce,) you will find that it has a fractal dimension of 0.9, or so (which is a 1/f power spectrum distribution.) What this means is that the market environment is only forcastable to 53% accuracy, or so, 6 months out-eg., the organization has about 6 months to adapt and learn a new market paradigm, or the organization will not be competitive in the market place. (If this is true, then annual MBO may not be suitable in such situations-and a rolling management methodology may have more specific application.) Note that, if this is true, then we would have a kind of formal basis for the learning organization concept. For what it is worth, just some thoughts, John BTW, while we are on the subject of game-theoretic solutions to management issues, a closely related works by Kenneth Arrow states that there is no logical methodology to determine priorities in a group of people, (the so called Impossibility Theorem.) Since executive management (a personal observation) is chartered with such things, and usually the priority determination process breaks down into a set of defection strategies (eg., parochial issues,) is there a formal relationship between the determination of priorities in a group and the iterated prisoner's dilemma? Just a curious thought ... -- John Conover, john@email.johncon.com, http://www.johncon.com/

Copyright © 1994 John Conover, john@email.johncon.com. All Rights Reserved. Last modified: Fri Mar 26 18:58:01 PST 1999 $Id: 941216133640.9245.html,v 1.0 2001/11/17 23:05:50 conover Exp $