🧸 Jiaxi Zhou
Projects
Projects / Featured

🕸️ Modeling Role-Differentiated Influence in Structured Social Systems

From structural influence to differentiated intervention strategy
Social systemsNetwork analysisLongitudinalLeadershipIntervention strategy

A unified longitudinal framework for detecting leadership, separating prestige and norm pathways, and modeling multilevel spillover for intervention design.

Decision: Who to target first, and where influence actually shifts outcomes.

Method: Role-differentiated longitudinal modeling across leaders, norms, and spillover pathways.

Outcome: Timing-sensitive, structure-aware interventions instead of one-size-fits-all averages.

Influence as structure: central and bridge users across communities with decision pipeline
Influence as structure. Central and bridge users create different intervention leverage across communities.
Method pipeline from relational nominations to decision translation
Method pipeline. Relational nominations → directed networks → role identification → multilevel modeling → decision translation.
Leader prestige vs norm climate pathways
Leader vs norm pathway. Separate prestige-based leader effects from descriptive norm climate effects.
Spillover cascade across communities
Spillover cascade. Bridge roles carry influence across clusters, shaping where interventions scale.

Core positioning

I build and test models that identify who shapes collective outcomes in real-world networks — and how different influence roles produce distinct downstream effects over time.

Informal leadership detection Prestige vs norm pathways Cross-level spillover Decision translation

Rather than treating groups as homogeneous units, this work differentiates roles, mechanisms, and leverage points.

Core Framework

Relational nominations → Directed networks → Community detection → Role identification → Multilevel longitudinal modeling → Decision translation

This framework integrates:

  • Directed, weighted network construction
  • WalkTrap community detection (igraph, R)
  • Centrality-based leader identification
  • Multilevel random-intercept and random-slope modeling
  • Cross-level interaction analysis
  • Longitudinal design

The goal is not to visualize networks — it is to identify structural leverage points.

Main Case: Leadership Strategy and Behavioral Spillover

2-year longitudinal design | N = 2,155 | 220 peer groups

In this study, I modeled how informal leadership strategies shape group-level behavioral cascades over time.

Leaders were identified within detected peer communities using nomination centrality. I then examined how leader behavior predicted downstream member outcomes across social, academic, and psychological domains.

Key Findings

  • Leadership status conferred structured individual advantage.
  • Prosocial leaders generated positive spillover across members.
  • Aggressive leaders increased risk-related outcomes.
  • Prosocial leadership buffered aggressive influence pathways.

This demonstrated that influence is:

  • Role-dependent
  • Mechanism-sensitive
  • Interaction-specific
  • Structurally uneven

Targeting average users misses these differentiated pathways.

Evidence Anchor 1: Leadership Emergence and Network Dynamics

Journal of Youth and Adolescence, 2024

Leadership is not a static label.

Using longitudinal network modeling (SIENA), I examined how leadership ties evolve over time and which attributes predict leadership emergence.

Findings showed:

  • Leadership ties were fluid at the dyadic level but moderately stable at the individual level.
  • Leadership nominations exhibited reciprocity and hierarchical “leader-of-my-leader” structures.
  • Higher social competence and aggression predicted leadership emergence; shyness predicted lower likelihood.
  • Academic performance and loneliness did not significantly drive leadership selection.

This work established that leadership roles are dynamically constructed within structured networks, not administratively assigned.

Evidence Anchor 2: Leader Prestige vs Norm Climate Effects

Journal of Research on Adolescence, 2026

In a separate longitudinal dataset (N = 2,450; 238 groups), I disentangled the effects of:

  • Leaders' academic achievement (prestige pathway)
  • Nonleader members' average achievement (descriptive norm pathway)

Both predicted later academic and social competence.

However, the pathways diverged:

  • Leaders' achievement reduced externalizing behaviors.
  • Nonleaders' achievement reduced peer victimization and internalizing symptoms.

This demonstrated that prestige-based influence and descriptive norm climate operate through parallel but distinct mechanisms.

Integrated Insight

Across studies, influence in structured systems is:

  • Unevenly distributed
  • Role-differentiated
  • Multi-level
  • Mechanistically distinct

Leadership prestige shapes behavioral regulation. Descriptive norm climate shapes socioemotional safety. Influence modeling must therefore distinguish:

  • Emergence (who becomes influential)
  • Role differentiation (how roles differ)
  • Spillover dynamics (how behavior cascades)

This program of work moves beyond global group averages toward structural influence modeling.

Organizational Translation

This framework can inform:

  • Informal leadership detection within teams
  • Influence-aware segmentation strategies
  • Culture interventions targeting high-leverage actors
  • Behavioral risk mitigation
  • Multilevel evaluation of intervention impact

Rather than optimizing for average behavior, this approach identifies who shifts the system.

Tools & Methods

  • R (igraph, network analysis)
  • SIENA (longitudinal network dynamics)
  • Multilevel SEM (Mplus)
  • Directed weighted graph modeling
  • Longitudinal cross-level inference

Why This Work Matters

In complex social systems — classrooms, organizations, communities, digital platforms — behavior is not evenly distributed. Understanding who drives norms, how influence propagates, and where intervention has leverage is central to effective decision-making.

This work provides a structured, empirically validated approach to modeling influence roles in real-world networks.