CIF: Small: Statistical Signal Processing of Social Networks with Behavioral Economics Constraints

Social networks are ubiquitous. There is strong motivation to understand important sociological phenomena in social networks by constructing novel engineering models, algorithms and analysis. This project contributes to fundamental research in understanding the interaction between statistical inference, human decision making and information flow in social networks.

The research addresses three interrelated themes. The first theme develops novel algorithms for detecting a change in the strategies of human decision makers that share information over a social network, for example, identifying reactive behavior to viral content on Twitter. The second theme models and analyzes how humans interacting over a social network can result in sociological phenomena such as the glass-ceiling effect and segregation.

The third theme studies efficient polling algorithms in large-scale social networks where only a fraction of nodes can be polled to determine their decisions. For example, which nodes should be polled to achieve a statistically accurate estimate of sociological phenomena such as the emergence of the glass-ceiling effect? The theoretical claims and findings in this project will be validated via extensive analysis of real-world social network datasets.

This project develops novel engineering models, algorithms and analysis to understand the interaction between statistical signal processing, behavioral economics (human decision making) and network science (information flow in social networks). The project conducts fundamental research in three interrelated themes. The first theme studies multi-agent information fusion and change detection with behavioral economics constraints (anticipatory decision making and rational inattention); the goal is to understand how information fusion is achieved among human decision makers.

The second theme investigates how sophisticated agents interacting over a social network give rise to the important sociological phenomena of the glass-ceiling effect and segregation. The glass-ceiling effect refers to the barrier that keeps certain groups from rising to influential positions in society, regardless of their qualifications. In a social-network context, the investigator explores how individual traits like preferential attachment and homophily leads to the glass-ceiling effect.

The third theme studies the design of statistically efficient network polling algorithms. In large-scale social networks, only a fraction of nodes can be polled to determine their decisions. Which nodes should be polled to achieve a statistically accurate estimate of sociological phenomena such as the glass-ceiling effect?

Some nodes may be reluctant to reveal their true opinion. This may lead to incorrect polling estimates. How can this be compensated for?

The theoretical claims and findings in this project will be validated via extensive analysis of real-world social-network datasets.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.