IPAN Tasks

The research in this CCRI will investigate social, cognitive and information models to support interactive search and query, visualization, and experimentation. From these models and experiments, theories will be validated and networks analytical processing methods will emerge to support the social network of decision makers. The first two tasks support the intelligence analyst, and the second two tasks investigate issues related to the tactical network of decision makers.

P1: Interactive Search and Query over Socio-Information Networks (Lead: M. Srivatsa (IBM))

Searching distributed information sources—both data repositories and human experts—is critical for decision making. The key philosophy behind expert search is that it is sometimes more efficient to ask an available human expert than iterating through search engines—especially for complicated information needs. This task supports efficient queries over distributed socio-information networks, which not only retrieve relevant information but also identify people (both individuals and groups) who have expertise related to the query. In doing so, the network considers the availability (or the lack thereof) of information sources and experts in conjunction with the delay tolerance of the query to deliver timely, high quality answers to decision makers.

P2: Modeling Scalable Recommendation of Credible Information Sources (Lead: J. O’Donovan (UCSB))

This task addresses the problem of recommending credible content across composite networks of social media and a range of other communication data. Practical limitations of information analyst s’ attention reinforce the need for automated assessments of credibility. Thus, the task explores the limitations, potential synergies, and other theoretical boundaries between automated credibility analysis algorithms and credibility assessments made by human analysts. Specifically, the integration of scalable credibility analysis tools, human-factors approaches, and cognitive modeling methods into an interactive visual interface for human analysis enable experimental mechanisms for cognitive assessment of analyst interaction with data from a number of credibility models.

P3: Online Laboratory to Study Large Group Network Performance (Lead: D. Lazer (NEU/Harvard))

This task performs a number of experiments to study how information processing and network structure support effective group decision making. While many experiments appear in the team literature, they largely neglect the team dimension because of (1) the difficulty in collecting dynamic network data, and (2) the cost and logistical challenges of running experiments with groups. These difficulties are especially true in the study of large groups; such study has been completely neglected (until very recently) in experimental research because of the impossibility of studying large groups in a laboratory setting. It is increasingly possible to use the Internet to recruit subjects and stage group experiments. This task designs and executes a set of online, group network experiments and provides a prototype infrastructure for the NS CTA to facilitate group network experiments. This infrastructure includes the development of a panel for human behavioral research, as well a set of templates that will greatly reduce the development of other group network experiments.

P4: Emergence of Communities and Leadership in Response to Extreme Events (Lead: W. Wallace (RPI))

The emergence of communities in networks requires a set of “qualified” individuals (leaders) embedded into one or multiple communities. For a community to “activate” to accomplish a particular goal, its members must communicate with one another. This research has the following goals: (1) validate the discovery of emergent online communities; (2) deactivate the communication in (potentially adversary) communities before its completion (prior to their action); and (3) discover ways to facilitate the effective activation of friendly combatant or humanitarian communities. It studies the formation and evolution of online human communities on Twitter in response to extreme events. The task traces the behavioral intent that users exhibit online and their role and engagement in online communities. The theory of behavioral intent suggests that the intent to engage in a particular behavior is the best predictor of the behavior. However, other factors may affect deviations in behavior from original intent, such as the quality of the information received, lack of behavioral control, etc. To address this issue, surrogate behavioral indicators need to be explored as well as metrics for the quality of information. The task studies community evolution over time and its controlling factors. The evolution of these communities can be assessed through structural and statistical analyses of the communities over multiple time periods. Understanding community evolution during extreme events will facilitate discovering guidelines for effective activation of friendly combatant or humanitarian communities.