p.96 Mutual causality is another key feature of the self-organizing system. The proposal that current levels of X can be partially predicted by prior levels of Y and current levels of Y can be partially predicted by prior levels of X captures a significant portion of the dynamism of evolutionary process. In the context of knowledge networks this type of mutual causality is reified again in the relationship between knowledge (X) and people (Y). On one hand, once people decide to share their knowledge with others, the knowledge becomes independent of the provider, and becomes information circulated in the network that can be used as raw material to generate more knowledge. Although we cannot say for sure what type of new knowledge will be created, accumulation and distribution of information and knowledge are necessary for the creation of new knowledge. On the other hand, as people are the actual inventors of new ideas, pooling together the intelligence of better informed and motivated people can spin off much more new knowledge.

The third condition states that maintaining a self-organized state requires importing energy into the system. Following the law of requisite variety as first proposed by Ashby, the level of internal complexity of a system has to match the level of complexity of the external environment if it is to deal with the challenges posed by that environment.

p.148 He [Polanyi, also Granovetter] observed that many theorists treated social institutions as if they were independent entities that were unconnected to other social institutions. He called this the "undersocialized" (or unembedded) view of economic action and argued that it represented one pole of a continuum on which various social theories could be placed. The other end of the continuum was anchored by the notion of "oversocialized" (or overembedded), which focused on the fact that many human and organizational economic actions were theorized to be extensively if not completely constrained by the social structures in which they are located. This view did not grant individuals or organizations much autonomy.

p.184-185 Missing in the contagion model is the typical ebb and flow of messages through networks that typifies human communication. In human interaction messages containing differing ideas, values, and attitudes flow back and forth among people as they negotiate resolutions. The most typical outcome is modifications to the different positions each person held at the outset of the contagion process, modifications that influence both contaminators and the contaminated.

p.203 In summary, transactive memory theory demonstrates that domains of knowledge are distributed throughout human and technical repositories in knowledge networks.

p.204-205 The generative mechanism posited by cognitive consistency is that the number of transitive triads in which they are embedded influences an actor’s attributes. Thus, as we shall see below, according to cognitive consistency theory, an individual’s satisfaction at work is influenced by the extent to which the individual is embedded in a large number of transitive triad relations. If the relation is a friendship relation, cognitive consistency theory argues that individuals are satisfied when their friends are friends with one another.

p.207 ... consistency theories have also been utilized to address the ongoing debate about differences between actual and perceived communication. Freeman suggested that consistency theories offer a systematic explanation for differences between actual and self-report data on communication. He argued that individuals’ needs to perceive balance in observed communication networks help explain some of the errors they make in recalling communication patterns. Using experimental data collected by De Soto, Freeman found that a large proportion of the errors in subjects’ recall of networks could be attributed to their propensity to "correct" intransitivity, a network indicator of imbalance, in the observed network. [Reference is to Freeman, L.C. 1992. “Filling in the blanks: A theory of cognitive categories and the structure of social affiliation.” Social Psychology Quarterly, 55, 118-127.]

p.257 "Commensalism refers to competition and cooperation between similar units, whereas symbiosis refers to mutual interdependence between dissimilar units." [Quote is from Aldrich, H. 1999. Organizations Evolving. Sage. P. 298.]

p.293-294 The first problem is the fact that the vast majority of network research is atheoretical. One reason for this is that there are very few explicit theories of social networks. Another reason is that researchers are generally not cognizant of the relational and structural implications inherent in various social theories. Even research that does employ theory typically does so without much attention to the network mechanisms implicit in the theories.

A second problem with network research is that most scholars approach networks from a rather myopic, single-level perspective, which is reflected in the fact that almost all published research operates at a single level of analysis. Thus, they tend to focus on individual features of the network such as density. For the most part, researchers tend to ignore the multiple other components out of which most network configurations are composed, structural components from multiple levels of analysis such as mutuality, transitivity, and network centralization. Employing single levels of analysis is not inherently wrong; it is simply incomplete. Importantly, these components suggest different theoretical mechanisms in the formation, continuation, and eventual reconfiguration of networks. Typically, better explanations come from research that utilizes multiple levels of analysis.

The third problem centers on the fact that most network research focuses on the relatively obvious elementary features of networks such as link density and fails to explore other, more complex properties of networks such as attributes of nodes or multiplex relations. But the members of networks often possess interesting theoretical properties, which help to shape the configurations in which they are embedded, and networks are themselves often tied to other networks. Traditional analyses typically account for relatively simple, surface features of networks and ignore these more subtle and sophisticated structural characteristics inherent in many networks.

The final problem is that most network research tends to use descriptive rather than inferential statistics. The reason for this is that network relations contain inherent dependencies that typically do not exist in traditional attribute data. These dependencies invalidate the assumption of independent observations, which forms the basis for much of the traditional social science research on attributes.