AI and Big Data Reshape China’s Online Public Opinion Governance
In the digital era, where information flows faster than ever, the governance of online public opinion has become a critical component of national governance and social stability. As internet penetration deepens across China, public sentiment expressed online increasingly reflects societal moods, emerging crises, and collective demands. In this evolving landscape, big data and artificial intelligence (AI) have emerged as transformative tools, reshaping how governments, media, and institutions monitor, analyze, and respond to public discourse. A recent comprehensive study by Li Mingde and Kuang Yan from Xi’an Jiaotong University, published in the Journal of Beijing University of Technology (Social Sciences Edition), offers a timely and incisive examination of the dual-edged nature of these technologies in the realm of online public opinion management.
The research, titled Network Public Opinion Management in the Context of Big Data and Artificial Intelligence: Value, Risk and Path Exploration, provides a systematic analysis of the technological advancements driving modern yuqing governance, while also sounding a cautious note about the ethical, technical, and social challenges that accompany them. As AI systems grow more sophisticated and data collection becomes ubiquitous, the study underscores the urgent need for balanced, human-centered governance frameworks that harness technological power without compromising democratic values or individual rights.
The Technological Revolution in Public Opinion Monitoring
One of the most significant contributions of big data and AI lies in their ability to transform public opinion monitoring from a reactive, fragmented process into a proactive, comprehensive system. Traditional methods of tracking online sentiment—relying on manual content review or limited keyword searches—struggle to keep pace with the sheer volume and velocity of digital communication. Social media platforms, forums, comment sections, and messaging apps generate petabytes of unstructured data daily, much of which contains valuable insights into public concerns.
According to Li and Kuang, AI-powered monitoring systems now enable real-time, large-scale data acquisition across multiple formats—text, audio, video, and even emojis. Through natural language processing (NLP), machine learning, and sentiment analysis algorithms, these systems can identify not only what users are saying but also the emotional tone behind their words. This shift from content-only analysis to a “content + relationship + emotion” framework allows authorities to detect early warning signs of social unrest, track the spread of misinformation, and understand the underlying psychological drivers of online discourse.
For instance, during the 2017 Jiuzhaigou earthquake in Sichuan, an automated reporting system developed by the China Earthquake Networks Center generated a detailed news article within 25 seconds of the event. The system pulled together geographical data, population density maps, and nearby settlements to produce an accurate and timely report. This case exemplifies how AI can accelerate official communication, reduce response delays, and counteract the spread of rumors during emergencies.
Moreover, AI-driven network analysis tools can map the structure of online conversations, identifying key influencers, echo chambers, and viral pathways. By visualizing the social network dynamics of information diffusion, policymakers gain a clearer picture of how narratives evolve and which nodes are most influential in shaping public perception. This level of granularity was previously unattainable with conventional monitoring techniques.
Enhancing Predictive Capabilities and Decision-Making
Beyond monitoring, AI and big data have significantly enhanced the predictive and decision-support capabilities of public opinion governance. The study highlights how machine learning models, including artificial neural networks and virtual reality-based simulation systems, can forecast the trajectory of events based on historical patterns and real-time data streams.
These predictive models allow authorities to anticipate potential crises before they escalate. For example, by analyzing fluctuations in sentiment intensity, keyword frequency, and user engagement metrics, AI systems can flag emerging issues—such as labor disputes, environmental protests, or consumer complaints—enabling preemptive interventions. In crisis management scenarios, cognitive AI systems can simulate the outcomes of different policy responses, allowing decision-makers to test strategies in silico before implementing them in the real world.
This predictive power is not limited to government actors. Media organizations are increasingly adopting AI to guide editorial decisions, optimize content distribution, and tailor messaging to specific audience segments. Personalized content delivery, powered by recommendation algorithms and user profiling, ensures that information reaches the right people at the right time, increasing the effectiveness of public communication campaigns.
However, the authors caution that predictive analytics should not be mistaken for deterministic forecasting. While AI can identify correlations and trends, it cannot fully account for the complex, often unpredictable nature of human behavior. Overreliance on algorithmic predictions risks creating a false sense of certainty, potentially leading to misguided policies or overreactions.
The Risks of Algorithmic Distortion and Misjudgment
Despite their advantages, AI and big data introduce new vulnerabilities into the public opinion ecosystem. One of the central concerns raised by Li and Kuang is the phenomenon of algorithmic distortion—where the very act of monitoring alters the behavior it seeks to observe. Drawing on Werner Heisenberg’s “uncertainty principle” from quantum physics, the researchers argue that digital surveillance systems can inadvertently influence online discourse.
For example, when social media platforms automatically highlight trending topics based on algorithmic calculations, they may amplify fringe or artificially inflated narratives. Users, in turn, engage more with these promoted topics, reinforcing their visibility and creating a feedback loop that distorts the true state of public opinion. In some cases, malicious actors exploit this mechanism by deploying bot networks to manipulate trending lists, creating false public opinion surges. If governance bodies rely solely on these algorithmically curated signals, they risk making decisions based on noise rather than genuine public sentiment.
Another major challenge lies in the accuracy of emotion detection algorithms. While NLP models can classify text as positive, negative, or neutral, they often struggle with sarcasm, irony, cultural nuances, and rapidly evolving internet slang. Chinese internet culture, in particular, is characterized by creative wordplay, homophonic puns, and context-dependent meanings—features that confound even the most advanced AI systems. As a result, sentiment analysis may misinterpret the intensity or direction of public emotion, leading to flawed assessments of situations.
The study cites instances where automated systems failed to detect subtle shifts in public mood, mistaking satirical criticism for genuine support or overlooking underground discussions happening in private chat groups. These blind spots are exacerbated by platform-specific restrictions; for example, content within WeChat Moments or private QQ groups remains largely inaccessible to external monitoring tools due to privacy settings and closed ecosystems.
The Rise of Information Cocoons and Content Manipulation
Perhaps one of the most insidious risks associated with AI-driven public opinion governance is the reinforcement of information cocoons. Coined by American legal scholar Cass Sunstein, the concept describes how personalized recommendation algorithms tend to filter out diverse viewpoints, exposing users only to content that aligns with their existing preferences. Over time, this creates isolated ideological bubbles where alternative perspectives are excluded, and extreme views are amplified.
In China’s digital environment, where platforms like Toutiao, Weibo, and Douyin employ sophisticated AI recommenders, the risk of polarization is particularly acute. The study notes that many content providers prioritize engagement metrics—click-through rates, dwell time, shares—over journalistic integrity or social responsibility. This incentivizes the production of sensationalist, emotionally charged, or entertainment-focused content, often at the expense of substantive public discourse.
When such content dominates algorithmic feeds, it not only distorts public understanding of important issues but also undermines the authority of mainstream media. Citizens immersed in these echo chambers may become skeptical of official narratives, perceiving them as out of touch or manipulative. This erosion of trust complicates efforts to guide public opinion in a constructive direction.
Additionally, the rise of automated content generation—such as AI-powered writing bots—introduces new risks of misinformation. While these tools can produce news reports quickly and efficiently, they lack the editorial judgment and ethical oversight of human journalists. Without proper safeguards, they may inadvertently propagate rumors, fail to verify sources, or be hijacked by bad actors to disseminate disinformation under the guise of legitimate reporting.
The problem is further compounded by widespread content recycling and plagiarism in the digital media space. Many platforms republish or “rewrite” existing articles without adding original insight, diluting the quality of public discourse and making it harder for audiences to distinguish credible information from low-effort aggregation.
Data Ethics, Privacy, and Monopolization Challenges
Underpinning all these technical and operational challenges are deeper concerns about data ethics, privacy, and corporate control. The study emphasizes that the foundation of AI-driven public opinion governance is vast amounts of personal data—search histories, location records, social interactions, and behavioral patterns. While this data enables granular analysis, its collection and use raise profound ethical questions.
Who owns user-generated data? Do individuals have the right to know how their information is being processed? Can companies indefinitely retain data collected for one purpose and repurpose it for another? These unresolved issues constitute what the authors call a “privacy ethics dilemma” in the age of big data.
Real-world incidents underscore the urgency of addressing these concerns. The Cambridge Analytica scandal involving Facebook revealed how personal data could be weaponized for political manipulation. In China, cases of data misuse—such as price discrimination based on user profiles (“big data price discrimination”) or unauthorized data sharing—have sparked public backlash and regulatory scrutiny.
Even more troubling is the concentration of data power among a few dominant tech firms. Citing a 2019 report by Renmin University’s Network and Mobile Data Management Lab, Li and Kuang note that the top 10% of data collectors control 99% of permission-based user data. This extreme imbalance creates a de facto data oligopoly, where a handful of companies wield disproportionate influence over the digital public sphere.
Such monopolization not only threatens competition but also hampers effective public opinion governance. When critical data is siloed within private platforms, government agencies and public media struggle to access comprehensive, interoperable datasets. This fragmentation limits the ability to form holistic views of societal trends and weakens evidence-based policymaking.
Toward a More Balanced and Human-Centered Governance Model
Recognizing these multifaceted challenges, the study proposes a series of strategic recommendations aimed at optimizing the role of AI and big data in public opinion governance. Central to their vision is the integration of technological innovation with human judgment, ethical principles, and institutional reform.
First, the authors advocate for a hybrid approach that combines big data analytics with traditional “small data” methods. While big data excels at identifying broad patterns, small data—such as in-depth interviews, expert analysis, and qualitative research—provides contextual depth and causal insight. By merging these two paradigms, governance systems can avoid the pitfalls of “big data arrogance,” the mistaken belief that correlation equates to causation.
Second, they emphasize the importance of interdisciplinary collaboration. Effective public opinion governance requires expertise not only in computer science and data engineering but also in journalism, sociology, psychology, and political science. Building cross-disciplinary teams ensures that technical solutions are grounded in social reality and aligned with public values.
Third, the study calls for stronger legal and regulatory frameworks to govern data use. While China’s Cybersecurity Law and subsequent regulations have established basic protections for user data, the authors argue that these rules need to be refined and enforced more rigorously. Clear standards for data ownership, consent, transparency, and algorithmic accountability are essential to prevent abuse and build public trust.
Fourth, the researchers stress the need to cultivate a new generation of fuhe xing professionals—individuals who possess both technical proficiency and a deep understanding of media, ethics, and public affairs. Universities and training institutions should develop specialized curricula that blend data science with communication studies, political theory, and moral philosophy.
Finally, the paper highlights the role of platform self-regulation and public education. Media organizations must uphold professional ethics, resist the temptation to chase clicks, and serve as gatekeepers of truth. At the same time, citizens need to develop greater media literacy and algorithmic awareness—understanding how recommendation systems work, recognizing bias in automated content, and critically evaluating online information.
Conclusion: Navigating the Future of Digital Governance
As Li Mingde and Kuang Yan conclude, the integration of big data and AI into public opinion governance represents both an opportunity and a responsibility. These technologies offer unprecedented capabilities to understand and respond to the pulse of society, but they also carry significant risks if deployed without caution, oversight, or ethical guidance.
The path forward lies not in rejecting technology, but in mastering it with wisdom and foresight. By fostering collaboration between technologists, policymakers, journalists, and civil society, China can build a more resilient, transparent, and inclusive system of digital governance. In doing so, it can set a global example for how nations can harness the power of AI not to control public opinion, but to better serve the public good.
Network Public Opinion Management in the Context of Big Data and Artificial Intelligence: Value, Risk and Path Exploration
Li Mingde, Kuang Yan
Journal of Beijing University of Technology (Social Sciences Edition)
DOI: 10.12120/bjutskxb202106001