The phenomenon of opinion polarization on social media has attracted significant scholarly attention. However, existing studies mostly focus on either user behavior or technological factors in isolation, offering limited insight into the deeper, systemic mechanisms behind its emergence. This paper examines the interactive relationship between users’ subjective preferences and algorithmic recommendations, analyzing their synergistic role in driving polarization. The findings suggest that opinion polarization is not driven by a single linear factor but rather emerges as a dynamic outcome shaped by the interplay between external technological forces and internal cognitive motivations. Individuals, influenced by psychological mechanisms such as cognitive dissonance and confirmation bias, tend to seek out information aligned with their preexisting positions. Algorithmic recommendations further reinforce this tendency by constructing increasingly homogenous information environments, thereby intensifying echo chambers and filter bubbles. Based on these insights, the paper recommends that platform algorithms be optimized to incorporate mechanisms for diverse information exposure and improved visibility of heterogeneous content. Furthermore, enhancing users’ information literacy, fostering rational expression, and cross-group dialogue are proposed as essential strategies to mitigate the adverse effects of opinion polarization on the public opinion ecology.
Zijun Zhang (Wed,) studied this question.