Research on the Impact of Algorithmic Bias in Generative Artificial Intelligence on Public Opinion Ecology from the Perspective of Media Ecology

Abstract

With generative artificial intelligence (AI) deeply embedded in the production and dissemination of public opinion, its inherent algorithmic biases are gradually reshaping the mechanisms of public opinion generation and the public’s cognitive structure. This paper, from the perspective of media ecology theory, uses content analysis and in-depth interviews to systematically compare the analytical texts of two generative AI models, Doubao and DeepSeek, in 30 highly discussed public opinion events. It also examines the impact of algorithmic biases on the public opinion ecosystem and individual cognition, incorporating the experiences of ordinary users. The study finds that generative AI exhibits a clear structural tendency in fact presentation, emotional expression, and value guidance. Its seemingly neutral expression, through repeated reinforcement, may influence the public’s cognitive framework and value judgments. This paper argues that generative AI has transformed from an information tool into a significant mediating force influencing the public opinion ecosystem, and the long-term impact of its algorithmic biases on public opinion urgently requires continued attention and reflection.