Location
Hilton Hawaiian Village, Honolulu, Hawaii
Event Website
https://hicss.hawaii.edu/
Start Date
3-1-2024 12:00 AM
End Date
6-1-2024 12:00 AM
Description
Social bots wield significant impact within social networks. Despite the widely recognized variations in individual responses to humans and bots, existing research has not thoroughly investigated the impact differences between human and social bots on individuals’ opinions. However, such differences are challenging to be estimated due to the presence of confounders introduced by homophily and the absence of counterfactual outcomes in observational network data. This study designs a counterfactual graph learning approach to accurately estimate causal effects, which exhibits superior performance in our simulations. The subsequent empirical results demonstrate that social bots yield a weaker influence than humans, and we further uncover diverse influential patterns of different types of opinions expressed by influence sources. Nevertheless, the impact difference is overestimated without applying our approach to control the confounders. Our research provides a practical approach and offers insights for stakeholders to scrutinize bots' impact from network perspectives.
Recommended Citation
Wu, Ziyue; Zhang, Yiqun; and Chen, Xi, "What if Social Bots Be My Friends? Estimating Causal Effect of Social Bots Using Counterfactual Graph Learning" (2024). Hawaii International Conference on System Sciences 2024 (HICSS-57). 5.
https://aisel.aisnet.org/hicss-57/dsm/data_analytics/5
What if Social Bots Be My Friends? Estimating Causal Effect of Social Bots Using Counterfactual Graph Learning
Hilton Hawaiian Village, Honolulu, Hawaii
Social bots wield significant impact within social networks. Despite the widely recognized variations in individual responses to humans and bots, existing research has not thoroughly investigated the impact differences between human and social bots on individuals’ opinions. However, such differences are challenging to be estimated due to the presence of confounders introduced by homophily and the absence of counterfactual outcomes in observational network data. This study designs a counterfactual graph learning approach to accurately estimate causal effects, which exhibits superior performance in our simulations. The subsequent empirical results demonstrate that social bots yield a weaker influence than humans, and we further uncover diverse influential patterns of different types of opinions expressed by influence sources. Nevertheless, the impact difference is overestimated without applying our approach to control the confounders. Our research provides a practical approach and offers insights for stakeholders to scrutinize bots' impact from network perspectives.
https://aisel.aisnet.org/hicss-57/dsm/data_analytics/5