Identifying Patterns and Signatures of Negative Behaviours in Networks.
Military / Coalition Issue
Adversarial groups seek to disrupt otherwise stable coalitions through negative social network ties, which work to fracture alliances. Through our work here, we now understand the local structures that are foundational and operative within such networks. This better enables us to anticipate and identify future disruptive ties. This work informs the cognitive dimension within information environment operations, which can be used to protect decision-making of coalition forces and disrupt decisions of adversaries.
Core idea and key achievements
Much research has been conducted on social networks comprised of positive ties, while relatively little research has been conducted on their corollary, negative-tie social networks. Therefore, this work contributes to our understanding of how, at a foundational and structural level, negative-tie social networks differ from positive-tie social networks.
We use Exponential Random Graph Models (ERGMs) to statistically model these negative-tie and positive-tie social networks, enabling us to discover and specify the precise mechanisms that contribute to the development of such negative-tie social networks. Specifically, we find that both positive-tie and negative-tie social networks contain more reciprocated dyads than expected by random chance. In contrast, we find that positive and negative networks differ in two key ways: triadic closure defines positive-tie networks only, while degree distributions are heavily skewed within negative-tie networks only. These constitute unique structural patterns that can be identified in new networks, enabling us to detect the inception of negative ties and their effects within otherwise stable coalitions and alliances.
Implications for Defence
Now that these underlying, structural signatures have been identified within positive and negative networks, novel networks can be analysed with these precise signatures in mind. They will indicate the presence of benevolent and malevolent actors and ties within networks, contributing to the stability of coalitions as action can be taken against any negative, hostile actors and ties.
Readiness & alternative Defence uses
Social Network Analysis is needed to investigate these structural signatures in novel networks, while our work here serves as the standard against which new analyses can be compared. As we have identified, triadic closure and degree distribution can now be utilized to investigate to what extent actors and ties comprise negative ties within that network.
Resources and references
Key related work:
- Cassie McMillan, Diane Felmlee and James Ashford “Reciprocity, transitivity, and skew: Comparing local structure in 40 positive and negative social networks.” American Soc Assoc. 2021
- Diane Felmlee, Cassie McMillan and Roger Whitaker “Dyads, Triads, and Tetrads: A Multivariate Simulation Approach to Uncovering Network Motifs in Social Graphs.” Applied Network Science.
- McMillan, Cassie, and Diane Felmlee. “Beyond dyads and triads: a comparison of tetrads in twenty social networks.” Social Psychology Quarterly 83, no. 4 (2020): 383-404.
- Cassie McMillan, Diane Felmlee, James Ashford and Emma Jayes “A Comparison of Local Structure in Positive and Negative Networks”
- Felmlee, Diane, Cassie McMillan, Don Towsley, Kun Tu, Roger Whitaker, and Gavin Pearson. “Social Network Motifs: A Comparison of Social Groups.”
- Felmlee, Diane Felmlee, Cassie McMillan, Roger Whitaker, Mudhakar Srivasta, Cheryl Giammanco and Emma Jayes “Identifying Social Network Patterns with Exponential Random Graph Models”
IBM UK, IBM US, Cardiff University, ARL, PSU