Complex Adaptive Systems

by Yaniv Altshuler

As mentioned in the previous post, social trading networks are a perfect example of complex systems, as their essence is directly derived from the collection of local interaction between the various traders. However, social trading networks also make a perfect example for a certain sub-category of complex systems, called Complex Adaptive Systems (CAS).

Complex Adaptive Systems are a special case of complex systems. They are complex in that they are dynamic networks of interactions and relationships not aggregations of static entities. They are adaptive in that their individual and collective behavior changes as a result of experience. For example, the behavior of many of a trading network’s users (and even the collective behavior of the network itself) can be altered by relatively slight stimuli that are induced to it.

To illustrate this, let us imagine some trading network that comprises 100,000 traders. Originally, each trader advocates a different trading strategy, and the collective behavior of the network resembles a heterogeneous “swarm” of relatively low internal consistency. Now, let us imagine that on a bright Monday morning user by the name Trading_Magician76 joins the network, and start trading. Amazingly, for the following few weeks the user demonstrates a striking ability to predict the markets, resulting in substantial revenues. Exposed to the rest of the network’s traders, user Trading_Magician76’s portfolio and open positions quickly gain growing attention by many of the network’s users, who start following these positions, adapting their behavior to this of user Trading_Magician76. Gradually, clear changes start to appear in the behavior of the network. Not only that the behaviors of many users are now influenced by their interactions with the new expert, but the network itself had transitioned into a network that displays high internal consistency, that is – the behaviors of its members are highly correlated (which can be tangibly shown through observations).

Another feature that is unique to Complex Adaptive Systems is the importance of high-level properties and measures such as self-similarity (having the high level system resembling a local part of itself, optionally in a recursive manner) and self-organization (having patterns of organization emerging throughout the system as a result of unplanned local interactions). A CAS can therefore be defined as a complex, self-similar collection of locally interacting highly adaptive agents.

In literature, Complex Adaptive Systems are often mentioned side by side with capabilities such as communication, cooperation, specialization, spatial and temporal organization, and reproduction. Specifically, this is true when looking at social trading networks – using information sharing to generate mutual increase in revenues (communication and cooperation), having individuals that specialize in market analysis, some that bring fresh updates from other forums and networks, and occasionally some that may possess some unpublished information (specialization).

On Networks – a specific type of systems – next time.