EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Optimal control strategies for belief formation in social learning networks

Yufei Liu, Bo Wang, Dongyan Sui, Siyang Leng

Journal of Physics: Complexity · 01.12.2026 · 7 (4) · 045015

DOI: 10.1088/2632-072x/aeb3e7 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Physics: Complexity, online erschienen am 03.11.2026. Verfasst von Yufei Liu und 3 weiteren Autor:innen. Die Arbeit nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Belief formation in social networks is shaped not only by peer interactions and noisy private observations, but also by persistent external interventions. To enable principled intervention design under uncertainty, we study the evolution of agents’ belief and formulate a continuous-time approximation of a discrete-time non-Bayesian social learning update, yielding a networked drift-diffusion model amenable to control analysis. Building on this representation, we propose two intervention mechanisms: direct influence, which perturbs agents’ observable evidence streams, and latent influence, which embeds a covert source into the interaction structure. For both mechanisms, we formulate a unified stochastic optimal control problem that steers the network toward prescribed subgroup targets while trading off tracking accuracy and control effort. Via Hamilton–Jacobi–Bellman analysis, we derive an explicit state-feedback optimal policy characterized by coupled Riccati differential equations. Experiments on real and synthetic networks reveal a trade-off between tracking accuracy and control effort: direct influence yields higher tracking accuracy, whereas latent influence requires lower realized control effort and exhibits greater robustness to environmental drift and observation noise. Further experiments demonstrate three-subgroup belief formation on a structured community network and show robustness under non-Gaussian innovations and intermittent control availability. These results provide quantitative guidelines for selecting intervention mechanisms in noisy, networked social learning systems.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Liu, Y., Wang, B., Sui, D., & Leng, S. (2026). Optimal control strategies for belief formation in social learning networks. Journal of Physics: Complexity, 7(4), 045015. https://doi.org/10.1088/2632-072x/aeb3e7

Vancouver

Liu Y, Wang B, Sui D, Leng S. Optimal control strategies for belief formation in social learning networks. Journal of Physics: Complexity. 2026;7(4):045015. doi:10.1088/2632-072x/aeb3e7

BibTeX
@article{Liu2026,
  author = {Liu, Yufei and Wang, Bo and Sui, Dongyan and Leng, Siyang},
  title = {Optimal control strategies for belief formation in social learning networks},
  journal = {Journal of Physics: Complexity},
  year = {2026},
  volume = {7},
  number = {4},
  pages = {045015},
  doi = {10.1088/2632-072x/aeb3e7}
}
Kostenlos starten →