Information cocooning-amplified by algorithmic filtering-poses complex challenges for emotional dynamics in online social networks. This study explores how algorithmically reinforced information cocooning shapes information diffusion and group emotional dynamics in online social networks. We propose a viewpoint-based network evolution model that simulates struc-tural transformations driven by user preferences. To model the hidden influence of personalized comment recommendations, we introduce the Hidden Comment Area Cocoon (H-CAC)-a novel higher-order structure that captures cocooning at the comment level. This structure is integrated into an emotion spreading mod-el, enabling the quantification of how cocooning affects collective sentiment. By defining Recommendation Accuracy (RA) as a tunable parameter, we systematically evaluate its impact on emo-tional volatility and polarization. Extensive simulations, validated with real-world data, reveal that while cocooning reduces content diversity, it can significantly enhance emotional resilience within groups. Our findings offer a new computational lens on the dual role of cocooning and provide actionable insights for designing emotionally stable, algorithmically governed social platforms.