In the context of an information-overloaded online environment, users are highly susceptible to subjective factors such as emotions and herd mentality, which can lead to a series of irrational behaviors, including the spread of rumors and herd malicious comments. These behaviors undermine the healthy development of the online ecosystem, and such behavior is particularly prominent among fan groups. This research integrates the General Aggression Model, Cognitive Dissonance Theory and Attribution Theory of Emotion, employing a combined approach of Structural Equation Modeling (SEM) and Fuzzy-set Qualitative Comparative Analysis (fsQCA). From the perspective of information cocoon, this research thoroughly analyzes the driving factors and configurational paths of herd malicious commenting behavior within fan groups, aiming to provide significant theoretical support for maintaining a healthy ecology on social media platforms. The SEM analysis results indicate that trait anger and information cocoon positively influence hostile attribution bias, while information cocoon and emotional commitment have a positive impact on cognitive dissonance. Hostile attribution bias, cognitive dissonance, and negative emotions all positively influence the willingness to engage in herd malicious comments. Additionally, negative emotions serve as a significant mediator between hostile attribution bias and herd malicious comment behavior, as well as between cognitive dissonance and herd malicious comment behavior. The fsQCA results indicate that both the information cocoon-enhanced and information cocoon-absent configurations trigger fans’ willingness to engage in herd malicious comment behavior. Based on these findings, social media platforms should introduce diversified information sources and reduce the effects of information cocoon to mitigate the spread of cognitive biases and negative emotions, thereby fostering a healthier online environment.