【报告摘要】The“filter bubble”effect has become a significant challenge in modern recommender systems, where personalized algorithms inadvertently confine users within narrow information ecosystems, reinforcing existing beliefs and limiting exposure to diverse perspectives. Consequently, users become progressively isolated from alternative viewpoints and conflicting information, posing significant threats to social cohesion, economic development, and individual intellectual development. The challenge is further complicated by the feedback loop inherent in recommendation systems, where user interactions continuously reinforce narrow preferences, making filter bubbles increasingly entrenched over time. An intuitive solution is to increase recommendation diversity. However, prioritizing content diversity often compromises recommendation accuracy, sacrificing user satisfaction and experience. This talk will share our research findings on mitigating the 'filter bubble' effect while maintaining user experience through novel approaches. Rather than forcing users to accept diverse recommendations, our methods guide them toward broader information exploration, demonstrating effective results in expanding user interests while preserving user satisfaction.