To appear in the Proceedings of the 4th International Conference on Applied Human Factors and Ergonomics (AHFE), San Francisco, CA (July 2012).
Analysts’ ability to understand and reason about complex organizations is often constrained by the limitations inherent in network analysis techniques and tools. Such tools are typically designed to work with homogeneous networks consisting of a single type of entity (e.g., a person) and a single type of relationship (e.g., “knows”). They also tend to be based on analysis techniques (e.g., graph-theoretic algorithms that measure centrality) that do not consider details about entities or relationships, or qualifying information about the data (e.g., whether the data is incomplete, stale, or from an unreliable source). Tools that support construction and analysis of heterogeneous networks often require operational analysts to have significant knowledge about network analysis—knowledge that analysts may not possess. In addition, most analysis techniques cannot be tailored to the social or cultural aspects of a particular situation. In many cases, analysts’ frustration with existing tools leads them to stop using the tools, and instead rely on ad-hoc or manual processes to understand networked information. We present several analysis methods that address these concerns to improve understanding of heterogeneous networks. These include extending traditional graph-theoretic network algorithms (e.g., centrality with respect to financial relationships while considering the uncertainty in the data); using causal influence models (CIMs) to let analysts map domain-specific reasoning to information about the network, and using a parallel coordinates plot that illustrates correlations across network data, algorithmic calculations, and analyses by other analysts.
For More Information
(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)