APPENDIX TO CHAPTER 6

The Corporation as a Complex

Adaptive System: Notes on the Brightline Model

Robert A. G. Monks

 

When we at LENS first began our work developing a dynamic, interactive model of the corporation as a CAS, we had visions of a computer game that could closely 'simulate' the reality of corporate life. Yes, corporations are complex, but we had no fears.  Computers today have vast amounts of memory and can handle extremely complex problems. Surely our computers could handle a full-scale corporation!

 

We were naïve. An agent-based model, we have come to realize, is not quite the same thing as a simulation, although the term "simulation" is often used to describe agent-based models. As David Lane has written, "The Santa Fe Institute modeling efforts described here all share one feature: they are not "simulations" … The very entities and relationships that appear in SFI models are highly abstracted from any world of direct experience. The abstractions, in turn, are based on presuppositions about what "worlds of experience" are like." These presuppositions, says Lane, are "far from the "common Sense" views of Western European/American cultural tradition that most of us have inherited," but they allow us to "construct interpretation of our own worlds of experience that open up new possibilities for effective action."[1]

 

This language certainly reflects the process we went through in creating our model — especially the part about new possibilities for effective action.

 

We also received helpful guidance from the writings of John Holland of the Santa Fe Institute, particularly his book Hidden Order: How Adaptation Builds Complexity.  Holland defines seven basic concepts (some "properties," some "mechanisms") necessary to modeling the complex adaptive system. In order of importance, these are:

 

·         Aggregation (property). This means putting similar things together and treating them as equivalent.

·         Tagging (mechanism). This is a function that makes aggregation possible — like a banner or flag.

·         Nonlinearity (property). Linearity means that we get a value for the whole by adding up the sum of its parts. Nonlinearity means that we may get less, the same, or more, depending on circumstances.

·         Flow (property). Flows are movements over a network made up of nodes (processors such as agents) and connectors.

·         Diversity (property). This concept says there are many niches, and if the inhabitant of a niche fails to adapt to change, some other inhabitant will fill that niche, ensuring diversity.

·         Internal models (mechanism). This is a means of anticipation, bringing the Cartesian res cognitans (the inside world) in contact with the Cartesian res externa (the outside world), so to speak.[2] (Another term is "schema," but that has a different meaning in math so we will not use it here.)

·         Building blocks (mechanism). This is the ability to break complexity down into parts and then recreate or create using the parts.

 

While substantially less ambitious in scope than originally intended, the final model we implemented does in fact exhibit all of these general characteristics.

 

VARIABILITY AND DEPLOYMENT

 

The Brightline model is intended for continual reuse in a number of settings.  It was written in Java, and in such a way that all agent attribute variables can be changed at run-time by the user.

 

The model has also been successfully deployed on a Web-based server, and thus made available via the Internet. Run-time users make a series of attribute variable settings and then run the model accordingly, with the results of each run presented both graphically, via behavioral pattern trendlines for each variable, and statistically, showing the relative changes in the corporation agents' market shares and corresponding shareholder responses.