社会化问答社区匿名性回答对用户参与的影响研究OACHSSCD

[Purpose/Significance]Social Question-and-Answer(Q&A)communities serve as pivotal platforms for global knowledge exchange and social interaction.While user participation is widely recognized as crucial for community vitality,research in both domestic and international literature examining how content features differentially influence user engagement under anonymous versus non-anonymous conditions remains limited.Most existing studies treat anonymity as a simple binary predictor,overlooking its role in shaping the nature of content and how the content engages users at the group level.This study addresses this gap by systematically comparing the distinct impact pathways of anonymous and non-anonymous answer characteristics on group-level user participation,thereby offering valuable insights for optimizing commu-nity interaction design and sustainability.[Method/Process]This research drew upon data from Zhihu,a leading Chi-nese Q&A community.The study employed a stratified sampling strategy to select 47 diverse discussion topics,which resulted in a final dataset of 54 151 answers,comprising 5 679 anonymous and 48 472 non-anonymous answers,created between March 2011 and June 2023.The analysis extracted multifaceted features encompassing information quality,including answer length,the number of unique words,novelty,readability,and creativity,alongside emotionality features such as emotional polarity,emotional word ratio,and modal particle ratio.To capture group-level user participation,the study innovatively quantified communicativity by calculating the network density of user interaction networks formed through reply relationships within comment sections.After evaluating multiple predictive models,the study selected and tuned the best-performing XGBoost model for anonymous and non-anonymous answers separately.Subsequently,the SHAP interpretable machine learning framework was applied to these optimized models to elucidate the specific contribu-tion and direction of influence for each feature.[Result/Conclusion]The analysis reveals two distinct influence pathways.Anonymous answers primarily foster user participation through emotional appeal,characterized by heightened emotionality and creativity.In contrast,non-anonymous answers effectively promote deeper engagement through quality-driven attri-butes,particularly text complexity and information relevance.By introducing network density as a measure of group-level participation and leveraging SHAP for model interpretation,this study transcends the limitations of traditional linear mo-dels and opaque machine learning algorithms in capturing complex nonlinear relationships.These findings contribute to the international discourse on online communities by elucidating the indirect role of anonymity,which operates not merely as a binary state but as a context that shapes content characteristics and their subsequent effects.The insights offer practical guidance for platforms to tailor content management and recommendation systems based on the distinct engagement dyna-mics of anonymous and non-anonymous contributions,ultimately fostering healthier and more vibrant knowledge ecosystems.