Finding influential subjects in a network using a causal framework

Area of Science:

• Network science
• Causal inference
• Public health

Background:

• Identifying influential individuals in networks is crucial for maximizing intervention impact in public health.
• Existing influence measures often rely on network structure or diffusion models, with an implicit causal assumption.
• The operative notion of influence in network research is often causal: identifying nodes for intervention to achieve maximal network-wide effects.

Approach:

• Define a causal notion of influence using potential outcomes framework.
• Review existing network influence measures, such as node centrality.
• Conduct simulation studies to compare causal influence with traditional centrality measures.

Key Points:

• Centrality measures may not always align with causal influence.
• The study provides a framework for understanding and measuring causal influence in networks.
• Simulation results show conditions under which centrality measures approximate causal influence.

Conclusions:

• A causal definition of influence is essential for effective network interventions.
• Understanding the assumptions behind centrality measures is critical for their application.
• The proposed causal framework offers a more robust approach to identifying influential individuals in network studies.