Network Structure Shapes Consensus Dynamics Through Individual Decisions

PNAS 2026 Network Science Consensus Dynamics Experiments Computational Modeling

// Quick Summary

Published in Proceedings of the National Academy of Sciences (2026)
How do shared narratives emerge in decentralized online networks?
Network structure shapes group consensus through individual-level decision mechanisms
Spatially-embedded networks produce echo chambers; fully-connected networks produce shared behaviors
Narrative complexity moderates these effects by engaging different social learning strategies
Findings generalize across human experiments and LLM-powered agent simulations

Overview

How do shared narratives emerge in decentralized online networks? Prior research using simplified group coordination tasks (e.g., face-naming) shows network structure shapes group consensus, but it is unclear whether these findings extend to more complex narrative coordination tasks that characterize everyday digital communication.

This study combines large-scale behavioral experiments in custom-built online social networks with LLM-powered agent-based simulations to test how network topology and narrative complexity interact to determine whether groups converge on shared meanings or fragment into echo chambers.

Overview of network experiment design — Experiment procedure and networked interaction tasks. Participants interacted across two network topologies — spatially-embedded ring-lattice and homogeneously-mixed — over 40 incentivized coordination trials following exposure to a narrative stimulus.

Key Findings

Spatially-embedded networks produce echo chambers; fully-connected networks produce consensus

Echo chamber onset across network conditions — Rewards and colormaps of hashtag responses across a single N = 20 run. Spatially-embedded networks (top) produce local clusters of coordinated behavior while fully-connected networks (bottom) drive group-wide consensus.

Narrative complexity moderates the effect of network structure on group consensus

Belief dynamics across network and content conditions — Onset of behavioral coherence during networked interaction. The effect of network structure on consensus is amplified under high-complexity narrative tasks (hashtag matching) relative to low-complexity tasks (face naming).

Narrative complexity shifts social learning strategy at the individual level

Social learning strategy dynamics — Proportion of each group adopting one of four decision strategies across 40 trials. Hashtag-matching groups sample new responses longer, while face-naming groups rapidly adopt self-consistent strategies.

Why It Matters

This work identifies the micro-level decision mechanisms through which network topology produces macro-level polarization or consensus. These findings have direct implications for how social platforms, media organizations, and public health communicators design information environments.

Platform design: Feed structures and connection topologies are not neutral — they systematically shape whether users converge on shared narratives or fragment into isolated belief clusters.
Content strategy: Narrative complexity interacts with network structure, suggesting that the cognitive demands of content modulate how topology shapes collective outcomes.
AI simulation: LLM-powered agent networks replicate key experimental patterns, opening a scalable path for testing intervention strategies before deploying them in real networks.

Citation

Priniski, J.H., Linford, B., Hirschmann, A., Venumuddala, S.K., Morstatter, F., Rodriguez, N., Brantingham, P.J., & Lu, H. (2026). Network structure shapes consensus dynamics through individual decisions. Proceedings of the National Academy of Sciences U.S.A., 123(2), e2520483123. doi:10.1073/pnas.2520483123