Prevention and Control of Public Opinion Polarization Risk under the Algorithmic Transformation of News Dissemination: Deconstruction of Generation Mechanism and Design of Intelligent Guidance Scheme

Algorithm-driven news dissemination has significantly improved information distribution efficiency but has also intensified the risks of information homogenization and public-opinion polarization. To address the formation of polarized opinion ecosystems caused by filter bubbles and echo-chamber effects, this study proposes an intelligent governance framework for algorithmic public-opinion environments. Semantic clustering and sentiment-polarity recognition techniques are employed to analyze the generation mechanism of polarization using nearly 10,000 algorithm-recommended news items collected from five mainstream platforms. Results indicate that the sentiment extremism index of algorithmically recommended content is 42.3% higher than that of manually edited content. An intelligent guidance framework integrating diversified recommendation strategies, user cognitive balancing models, interpretable algorithm interfaces, and reinforcement-learning-based adaptive feedback is subsequently developed. The proposed approach provides effective support for rational information dissemination and offers methodological references for communication-network analysis, sentiment computing, and intelligent information governance systems.