Dynamic Information Dissemination Model Incorporating Non-Adjacent Node Interaction

Xinyu Li, Jinyang Huang, Xiang Zhang, Peng Zhao, Meng Wang, Guohang Zhuang, Huan Yan, Xiao Sun

Abstract

Describing the dynamics of information dissemination within social networks poses a formidable challenge. Despite multiple endeavors aimed at addressing this issue, only a limited number of studies have effectively replicated and forecasted the evolving course of information dissemination. In this paper, we propose a novel model, DM-NAI, which not only considers the information transfer between adjacent users but also takes into account the information transfer between non-adjacent users to comprehensively depict the information dissemination process. Extensive experiments are conducted on six datasets to predict the information dissemination range and the dissemination trend of the social network. The experimental results demonstrate an average prediction accuracy range of 94.62% to 96.71%, respectively, significantly outperforming state-of-the-art solutions. This finding illustrates that considering information transmission between non-adjacent users helps DM-NAI achieve more accurate information dissemination predictions.

Introduction

Traditional social network models, such as the Independent Cascade (IC) and Linear Threshold (LT) frameworks, typically assume that information only spreads between direct "friends" or followers. However, in the modern era of algorithmic feeds and "explore" pages, users frequently encounter information from strangers. This paper introduces DM-NAI, a novel model that accounts for interactions between non-adjacent users and the dynamic evolution of user attitudes to more accurately forecast how information ripples through a network.

The DM-NAI Model

The proposed model rests on three primary pillars designed to capture the complexity of real-world social interactions:

  1. Attitude Distribution Similarity: Instead of a binary state, each user is represented by a distribution of attitudes (positive, neutral, negative, or unknown) across multiple topics. Influence is calculated based on the similarity of these perspectives.
  2. Non-Adjacent Interaction (NADJ): This core innovation allows information to jump between users who are not directly connected, provided their attitude similarity exceeds a specific threshold ($\tau$).
  3. Dynamic Attitude Change: The model introduces "attitude perseverance" ($A_v^i$), a metric of user stubbornness. As users encounter information, their perseverance updates, capturing the "echo chamber" effect where repeated exposure to aligned views reinforces a stance.
Figure 1: Illustration of attitude distributions across multiple topics within a social network.

Experimental Results

The researchers evaluated DM-NAI using six real-world microblog datasets. The model was compared against several state-of-the-art baselines, including IC, Time-aware IC (TIC), Emotion-based IC (EMIC), and Extended IC (EIC).

  • Superior Accuracy: DM-NAI achieved an average prediction accuracy (AUC) between 94.62% and 96.71%, significantly outperforming the classic IC model (~60%) and advanced models like EIC (~84%).
  • Trend Prediction: The model successfully replicated the dissemination trends observed in real-world data, tracking the number of activated users over time with high precision.
  • Stance Prediction: Beyond simple activation, the model predicted user stances with approximately 77-79% accuracy.
Figure 3: Comparison of DM-NAI against baseline models in predicting the number of activated users over time across three datasets.

Key Insights

A significant finding of the study is that non-adjacent interactions are critical for predictive accuracy; including them caused a substantial jump in performance compared to neighbor-only models. Qualitatively, the experiments also showed that as information spreads, users initially holding supportive or neutral stances often shift toward negative attitudes regarding public opinion topics.

Conclusion and Future Work

By modeling "stubbornness" as a dynamic variable and accounting for the influence of non-neighbors, DM-NAI provides a more realistic depiction of information flow. While the model demonstrates high accuracy, future work may explore its computational scalability for massive networks and the integration of platform-specific algorithmic biases. The authors suggest such models could eventually serve as early warning systems for adverse social events.

Summary updated

Abstract

Describing the dynamics of information dissemination within social networks poses a formidable challenge. Despite multiple endeavors aimed at addressing this issue, only a limited number of studies have effectively replicated and forecasted the evolving course of information dissemination. In this paper, we propose a novel model, DM-NAI, which not only considers the information transfer between adjacent users but also takes into account the information transfer between non-adjacent users to comprehensively depict the information dissemination process. Extensive experiments are conducted on six datasets to predict the information dissemination range and the dissemination trend of the social network. The experimental results demonstrate an average prediction accuracy range of 94.62% to 96.71%, respectively, significantly outperforming state-of-the-art solutions. This finding illustrates that considering information transmission between non-adjacent users helps DM-NAI achieve more accurate information dissemination predictions.
BibTeX
@misc{li2024dynamic,
  author = {Xinyu Li and Jinyang Huang and Xiang Zhang and Peng Zhao and Meng Wang and Guohang Zhuang and Huan Yan and Xiao Sun and Meng Wang},
  title = {{Dynamic Information Dissemination Model Incorporating Non-Adjacent Node Interaction}},
  year = {2024},
  eprint = {2403.13267},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2403.13267},
}
Figure