opening it up with Common Lisp
Book review: Darwinia
Summer reading: Spin
the Omnivoire's Delimma
the Golem's Eye
Dynamic Network Analysis (DNA) blends traditional Social Network Analysis (SNA) with multi-agent simulation, cognitive modeling and machine learning to produce a tool appropriate for changing and uncertain (probabilistic) environments. DNA is applicable, for example, when modeling terrorist organizations: the agents and their relations are partially known at best and intentionally misleading at worst. The network is also subject to constant change in makeup, linkages, resources and goals.
To handle such problems, Carley puts forth her own meta-matrix approach which seems to be no more than the injunction to examine multiple networks at once (e.g., people/people, people/knowledge, people/events, knowledge/knowledge, knowledge/needs, events/events, and so on) to create a whole. She claims that most traditional SNA metrics have little value in such large multi-faceted networks and that new metrics are therefore needed. One such metric is the cognitive load of agents in the network. This amalgamates interactions, coordination costs, events and learning / training.
Given s snapshot of such a matrix, one can then examine how it will change over time. For example, the people in the network may be born, die, move, and add and sever connections; the knowledge/resources network may expand with innovation, contract with amnesia and change with technical discovery; the relations in either network change from cognitive, social and political processes
DyNet is Carley's tool for analyzing networks in DNA terms. It very much a work in progress but seems promising and has been used in experiments to determine how different kinds of groups react to different kinds of isolation strategies.
Psychology and Sociology can be seen as another manifestation of the two poles of the Nature / Nurture debate: are we governed by who we are (our own agency) or by the roles we play (our social network). It's fairly obvious at this point that neither answer is accurate and that our agency and roles co-evolve (cf. Susan Oyama's work for a wonderful and deep discussion of this in terms of the phenotype / genotype distinction). DNA may be too affiliated with other connotations for the term to catch on, but the goal of modeling complex systems as agents with social roles that use resources, pass information and change their affiliations is, in my opinion, spot on.
Copyright -- Gary Warren King, 2004 - 2006