Abstract: The rapid development of intelligent technologies is propelling humanity toward an intelligent society, bringing about profound changes in production, lifestyle, organizational structures, and social order. These transformations have given rise to numerous new challenges, making the governance of intelligent society an urgent research agenda. While existing studies have addressed key risks associated with intelligent technologies—including algorithmic abuse, the digital divide, and issues of trust and fatigue—relatively less attention has been paid to the transformative impact of intelligent recommendation systems on public information exposure, cognitive patterns, and behavioral preferences. The widespread use of personalized recommendations often results in individuals being repeatedly exposed to homogeneous content, leading to the formation of information cocoons, exacerbating cognitive polarization, and increasing the risk of societal fragmentation. This presents a significant challenge for intelligent social governance. Traditional analyses of public cognition and attitude formation are often based on the rational actor model, assuming that individuals make decisions through cost-benefit calculations. However, such perspectives tend to overlook the psychological and emotional mechanisms involved in information processing. Recent studies from the fields of psychology and behavioral science have begun to explore how emotional preferences, identity, ideology, and personality traits influence the relationship between information exposure and cognitive polarization. Nevertheless, little attention has been given to the role of perceived external social environments, especially perceptions of societal threats. Psychological research indicates that perceived threats in the social environment significantly shape individual cognition and attitudes. In today’s digital landscape, where intelligent technologies dominate, threat-inducing and emotionally charged content is more likely to attract attention and be widely disseminated. This can amplify public anxiety and perceptions of insecurity, thereby intensifying cognitive polarization. Accordingly, this study seeks to bridge the gap by examining how threat perceptions and information exposure interact to influence public cognitive polarization. This paper introduces the perspective of external threat situations and explores how they, together with information exposure, jointly influence public cognitive polarization. Using the application of facial recognition technology in “pedestrian red-light violation exposure” as a case study, this research constructs a 2 (no threat situation vs. threat situation)×2 (homogeneous information vs. heterogeneous information) between-subjects experiment. A total of 1,281 participants were recruited online for the experiment. The results indicate that exposure to homogeneous information strengthens the public’s original views, thereby intensifying cognitive polarization, while exposure to heterogeneous information weakens the public’s original views, reducing cognitive polarization. Compared to the absence of external threat situations, threat situations significantly weaken the depolarization effect brought about by exposure to heterogeneous information. Additionally, for individuals of different genders and average annual incomes, threat situations play varying roles in the process by which information exposure influences polarization. The findings demonstrate that the external environment in which the public is situated, including threat perceptions caused by social unrest, economic crises, crime, disease risks, and unexpected events, may have a significant impact on changes in cognitive polarization during information exposure. This deserves further attention and research. Moreover, when the government addresses viewpoint polarization in cyberspace, it should not only promote the public’s exposure to more heterogeneous information that challenges their original views and encourages diverse thinking, but also pay attention to the external environment the public faces. This can help reduce public anxiety and threat perception, thereby enhancing the role of heterogeneous information in mitigating viewpoint polarization.