[Paper Review] Flipping Stance: Social Influence on Bot's and Non Bot's COVID Vaccine Stance.

Ng Lhx, Kathleen M. Carley|arXiv (Cornell University)|Jun 21, 2021

Misinformation and Its Impacts14 references4 citations

TL;DR

This paper proposes a social influence model to analyze stance flipping toward COVID-19 vaccines on Twitter from April 2020 to May 2021, incorporating agents' past tweets, network structure, and neighbor influence. The model achieves 86% accuracy in predicting stance changes, revealing that bots (53.7% of flip-floppers) require less social influence to switch stances compared to non-bots.

ABSTRACT

Social influence characterizes the change of opinions in a complex social environment, incorporating an individual's past stances and the impact of interpersonal influence through the social network influence. In this work, we observe stance changes towards the coronavirus vaccine on Twitter from April 2020 to May 2021, where 1\% of the agents exhibit the stance flipping behavior, of which 53.7\% are identified bots. We then propose a novel social influence model to characterize the change in stance of agents. This model considers an agent's and his neighbor's past tweets and the overall network structure towards a stance score. In our experiments, the model achieves 86\% accuracy. In our analysis, bot agents require lesser social influence to flip stances and a larger proportion of bots flip.

Motivation & Objective

• To understand how social influence drives stance changes on social media during the COVID-19 pandemic.
• To identify and quantify the role of bots versus non-bots in stance flipping behavior toward the coronavirus vaccine.
• To develop a novel social influence model that integrates individual history, network structure, and neighbor influence to predict stance shifts.
• To analyze the differential influence thresholds required for bots and non-bots to flip their stance on vaccine-related topics.

Proposed method

• The model computes a stance score based on an agent’s past tweets and the collective influence of their social network neighbors.
• It incorporates the overall network structure, including centrality and connectivity patterns, to weight social influence effects.
• A dynamic stance representation is learned using temporal tweet sequences and sentiment trends from April 2020 to May 2021.
• The model uses supervised learning to predict stance flipping events, with features derived from tweet content, user history, and network topology.
• It classifies agents as bots or non-bots using automated detection methods and compares their flipping behaviors.
• The model is trained and evaluated on a Twitter dataset, achieving 86% accuracy in stance flip prediction.

Experimental results

Research questions

• RQ1 What proportion of stance flipping behavior on vaccine-related topics is driven by bots versus non-bots on Twitter?
• RQ2 How does social influence from neighbors and network structure affect an agent’s likelihood to flip stance on the coronavirus vaccine?
• RQ3 Do bots require less social influence to flip stances compared to non-bots?
• RQ4 How does an agent’s past tweet history interact with network-level influence to predict stance changes?
• RQ5 What is the predictive accuracy of a social influence model that integrates individual behavior and network structure?

Key findings

• A total of 1% of agents exhibited stance flipping behavior toward the coronavirus vaccine on Twitter between April 2020 and May 2021.
• Of the agents who flipped stance, 53.7% were identified as bots, indicating a significant role of automated accounts in opinion shifts.
• Bots required significantly less social influence to flip stances compared to non-bots, suggesting higher susceptibility to network-level persuasion.
• The proposed social influence model achieved 86% accuracy in predicting stance flipping events.
• The model's performance highlights the importance of integrating individual history and network structure in modeling opinion dynamics.
• The findings suggest that bots are not only more prevalent in stance flipping but also more responsive to social influence than non-bots.