Foshan Launches AI-Powered Index to Quantify Online Positive Energy
In an era where digital narratives shape public perception more than ever, measuring the impact of positive messaging has remained a persistent challenge for governments and institutions worldwide. Traditional metrics—such as page views, shares, or likes—fail to capture the nuanced influence of content that aligns with civic values, policy advocacy, or social cohesion. Enter Foshan, a dynamic city in Guangdong Province, China, which has pioneered a novel solution: the Foshan Online Positive Energy Index. Developed under the guidance of the Foshan Cyberspace Affairs Commission, this system leverages media big data, natural language processing, and machine learning to systematically track, analyze, and quantify how effectively local government units disseminate constructive narratives across the digital ecosystem.
Unlike generic sentiment analysis tools or social media dashboards, the Foshan Index is purpose-built for public-sector communication evaluation. It doesn’t just count mentions—it contextualizes them. The system evaluates not only where a story appears but also the editorial authority of the outlet, the prominence of placement (homepage vs. subpage), and the semantic alignment with predefined “positive energy” criteria rooted in national policy frameworks and socialist core values. This approach transforms qualitative communication efforts into a standardized, comparable, and actionable metric—offering a rare empirical lens into the otherwise elusive domain of state-led digital narrative building.
The genesis of the index lies in a practical dilemma faced by many municipal departments: after organizing community initiatives, policy rollouts, or public welfare campaigns, officials often struggle to assess their media footprint comprehensively. Coverage might appear on a provincial news app, a district WeChat official account, or a national portal—but tracking it manually across hundreds of platforms is inefficient and error-prone. Worse, search engines like Baidu often miss content published on mobile-native platforms or behind paywalls. The result? Incomplete reporting, skewed performance evaluations, and missed opportunities for strategic refinement.
To address this, the Foshan Cybersecurity and Informatization Association, led by Secretary-General Fan Chunbai, embarked on a two-year development journey culminating in a proprietary software system granted a national copyright certificate in November 2019. The index now monitors over 300 government entities—including 71 municipal, 202 district-level, and 32 township or subdistrict offices—across a curated list of 253 verified news sources. These range from core central media like People’s Daily Online and Xinhua Net to local community newspapers and official WeChat channels, all vetted for compliance with China’s Internet News Information Service Licensing requirements.
At the heart of the system is a multi-stage data pipeline. First, it deploys automated crawlers configured to specific content tags and API endpoints across diverse platforms—news websites, mobile apps, social media accounts, and content aggregators. Unlike blanket scraping, the system uses targeted collection rules: for instance, it may monitor the “Guangdong” section of Southern+ or the “Foshan” tag on Toutiao. This ensures relevance while minimizing noise.
Once raw content is ingested, the system performs structural parsing to extract titles, metadata, body text, and publication context. This is followed by semantic analysis. Using a dynamic keyword library—continuously refined through human-in-the-loop feedback—the algorithm identifies whether a piece qualifies as “positive energy.” Crucially, routine administrative activities such as internal meetings, policy briefings, or staff training sessions are explicitly excluded. Only content that addresses major policy implementations, public events, social issues, or crisis responses—and does so in a way that promotes civic unity, ethical behavior, or national values—is counted.
The classification logic employs Boolean combinations: for example, a valid article must contain a Foshan-related geographic term (e.g., “Foshan,” “Nanhai,” “Shunde”) AND a thematic keyword (e.g., “volunteer,” “civilization,” “disaster relief”) while NOT containing exclusionary terms like “internal notice” or “closed-door session.” This rule-based filtering is augmented by machine learning models trained on thousands of manually labeled examples, enabling the system to distinguish genuine public-interest reporting from perfunctory bureaucratic announcements.
But perhaps the most innovative aspect of the Foshan Index is its weighted scoring mechanism. Recognizing that not all media outlets carry equal influence, the system assigns tiered weights based on both the outlet’s institutional authority and the visibility of the article’s placement. Core central media (e.g., CCTV.com, China Radio International) receive the highest weight, followed by key central media (e.g., Guangming Net, China Youth Online), provincial outlets (e.g., Southern Net, The Paper), municipal platforms (e.g., Foshan Daily Online), district-level channels, and finally township-level sources. Moreover, an article featured on a homepage or in a push notification earns a higher score than one buried in a third-tier subcategory.
This weighting is further calibrated by administrative level. A township office securing coverage in People’s Daily Online—a rare feat—receives a disproportionately higher score than a municipal bureau achieving the same, acknowledging the structural asymmetry in media access. Such design choices reflect a deep understanding of China’s multi-layered governance-media ecosystem, where central media dominance coexists with hyperlocal digital outreach.
The index also accounts for multi-platform syndication. If a single story is republished across five different outlets—with one outlet featuring it three times in different sections—the system counts five “nodes” rather than one article. This captures the true propagation footprint, recognizing that repetition across channels amplifies message penetration.
After nearly two years of iterative refinement, the system now achieves over 90% precision in identifying relevant positive-energy content. This accuracy stems from a hybrid human-AI workflow. Initially, the model operated with a fixed rule set, casting a wide net but yielding many false positives. Over time, human reviewers flagged borderline cases—“provisional selections”—which were then fed back into the training loop. The model gradually learned to discern subtle contextual cues: for instance, differentiating a genuine community cleanup campaign from a staged photo-op, or recognizing when a policy explanation genuinely clarifies public concerns versus merely reciting bureaucratic jargon.
The results have proven remarkably aligned with real-world events. During major public health responses, natural disaster relief efforts, or national commemorative campaigns, spikes in the index closely mirrored actual civic mobilization and media attention. Units that actively coordinated with journalists, produced compelling visual content, or engaged citizens through interactive formats consistently ranked higher—validating the index as a proxy for effective public communication.
Beyond performance benchmarking, the index serves as a strategic compass. Monthly rankings are published internally, accompanied by diagnostic summaries for top performers. These insights help other departments emulate best practices—whether it’s timing releases to coincide with national themes, tailoring messages to specific demographic segments, or leveraging multimedia storytelling. In this way, the index fosters a culture of data-informed communication within the bureaucracy.
The initiative has drawn recognition beyond Foshan. Both the Guangdong Provincial Cyberspace Affairs Office and the Central Cyberspace Affairs Commission have acknowledged its innovation, citing it as a model for “scientifically guiding online public opinion” in line with President Xi Jinping’s vision of a “strong cyber nation.” By transforming abstract ideological goals into measurable KPIs, the system bridges the gap between political intent and operational execution.
Critically, the Foshan Index avoids the pitfalls of vanity metrics. It does not reward clickbait, sensationalism, or engagement farming. Instead, it incentivizes substance, alignment with public interest, and cross-platform coherence—qualities often absent in commercial social media algorithms. In doing so, it offers a counterpoint to the attention economy, proposing an alternative paradigm where digital success is measured not by virality, but by civic contribution.
From a global perspective, the project raises intriguing questions about the role of AI in democratic communication. While rooted in China’s specific governance context, its technical architecture—semantic filtering, source credibility weighting, multi-channel attribution—could be adapted to other public-sector communication challenges, from public health messaging to climate advocacy. The core insight remains universal: in an information-saturated world, the ability to measure meaningful impact—not just volume—is essential for responsible institutions.
Looking ahead, the Foshan team plans to integrate real-time sentiment feedback from citizen comments and expand monitoring to short-video platforms like Douyin (TikTok’s Chinese counterpart), where youth engagement is increasingly concentrated. They also aim to develop predictive analytics—using historical index trends to forecast optimal communication windows for upcoming policy launches.
In sum, the Foshan Online Positive Energy Index represents a sophisticated fusion of public administration, data science, and media theory. It demonstrates how governments can harness AI not for surveillance or control, but for self-assessment, transparency, and service improvement. In an age of digital fragmentation, such tools may be indispensable for rebuilding shared narratives and fostering trust between institutions and citizens.
By Fan Chunbai, Foshan Cybersecurity and Informatization Association, Foshan, Guangdong 528000, China. Published in China Media Technology, 2021(08):29–31. DOI: 10.19483/j.cnki.11-4653/n.2021.08.006.