📄 social_science

Digital Media Art Consumption and Symbolic Domestication Among Chinese Otaku Using a Multimodal Machine Learning Approach

This study employs a multimodal machine learning approach to analyze Chinese social media data, revealing how specific digital art features like system narratives and Guochao integration drive symbolic domestication and consumption intentions among the Chinese Otaku community, while highlighting the moderating role of community co-creation and generational disparities.

Original authors: YINGTING ZHANG, KUN FENG

Published 2026-07-10

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: YINGTING ZHANG, KUN FENG

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the world of Chinese "Otaku" (fans of anime, games, and virtual idols) not as a quiet group of people hiding in their rooms, but as a massive, roaring digital city with hundreds of millions of residents. A new study by Zhang Yingting and Feng Kun decided to peek behind the curtain of this city to understand what makes these fans tick, what they buy, and why they feel so connected.

They didn't just ask people what they liked; they acted like digital detectives, using a super-smart computer brain (a machine learning model) to scan 81,457 text comments and 123,692 images from popular Chinese social media apps like Bilibili, Xiaohongshu, and Weibo. They looked at data from 2024 to early 2026 to find the secret ingredients of fan culture.

Here is what they discovered, explained without the heavy jargon.

The Big Myth They Busted: It's Not Just About "Cute"

For a long time, experts thought Japanese fans (and by extension, Chinese fans) were mostly obsessed with "Moe elements"—that's the fancy word for cute character traits like "tsundere" (grumpy but sweet) or "loli" (young-looking). The old theory suggested fans just collected these cute fragments like trading cards, ignoring the big story.

The paper argues this is wrong for Chinese fans.
Instead of just collecting cute pieces, Chinese Otaku are actually obsessed with "System Narratives." Think of it like this: If a Japanese fan might love a character just because they have big eyes and a specific catchphrase, a Chinese fan is more likely to love the entire rulebook of the world that character lives in. They want to know the logic, the progression rules, and the grand architecture of the universe. They aren't just looking at a picture; they are solving a puzzle.

The study found that while "cute" stuff is important, the deep dive into complex, logical systems (like cultivation levels in games or world-building rules in novels) is actually the stronger driver for what fans decide to buy and how deeply they get involved.

The "National Tide" Secret Sauce

The researchers also looked at something called "Guochao Integration." This is when traditional Chinese culture (like Hanfu clothing, Taoist philosophy, or ancient myths) gets mixed into modern anime and games.

They found that this isn't just about looking cool. It works like a two-step magic trick:

Seeing these traditional symbols makes fans feel proud of their culture ("Indigenous Cultural Identification").
That feeling of pride then makes them want to spend money on the product.

The study suggests that about 40% of the reason why fans buy "Guochao" products is because of this feeling of cultural pride. It's not just a visual gimmick; it's a deep emotional connection.

The Power of the "Co-Creation" Club

Another huge finding is about the fans themselves. The study calls this "Community Co-creation Intensity." This is when fans don't just watch; they make their own art, write fan fiction, or build mods.

The paper shows that when a community is super active in making their own stuff, it acts like a volume knob for the "Guochao" effect. If the community is quiet, the traditional cultural elements don't push sales as hard. But if the community is loud, creative, and making their own content, those traditional elements become much more powerful at driving fans to buy things. It's like a campfire: one person lighting a match is nice, but a whole group throwing wood on it makes a massive fire.

The Generation Gap: "Visuals" vs. "Stories"

The researchers noticed a clear split between two age groups, which they call Generation Z (born 1995–2009) and Millennials (born 1980–1994).

Generation Z is like a fast-food lover of visual culture. They are driven by immediate "Moe" visuals and the excitement of the community chat. They want the "cool factor" right now.
Millennials are more like slow-cooked stew lovers. They care more about the deep, logical "System Narrative" and the long-term story. They want to know the rules and the history.

Interestingly, even though Millennials care more about the story structure, Generation Z actually showed a stronger connection to "Guochao" (traditional culture) and national pride. It seems the younger generation, growing up during China's rise, has a very natural, instinctive love for these cultural symbols, even if they prefer fast-paced visuals over slow stories.

How Sure Are We?

The authors didn't just guess; they used a heavy-duty math model called XGBoost combined with a tool called SHAP to measure exactly how much each factor matters.

They found that 87.64% of the ratings in their data were 4 or 5 stars, showing these fans are generally very happy and engaged.
Their model was very good at predicting what fans would do, getting a score (R²) of 0.763 for purchase intentions. This means the factors they found (like System Narratives and Cultural Pride) explain a huge chunk of why fans buy things.
They also proved that looking at both text and images together was crucial. If they had only looked at text, they would have missed 285% more of the "System Narrative" clues that were hidden in the images (like game interface diagrams).

What This Means for the Future

The study suggests that if you want to make a hit game or anime for Chinese fans, you can't just slap on a cute character and hope for the best. You need to build a world with solid rules and logic. You need to weave in traditional culture in a way that makes fans feel proud, not just like it's a costume. And most importantly, you need to let the fans play, create, and talk, because their creativity is the fuel that makes the whole thing work.

The paper admits it has limits—it mostly looked at public social media and didn't check private chat groups, and it's a snapshot in time, not a movie of the future. But for now, it gives us a very clear, data-backed map of how Chinese Otaku think, feel, and spend their money.

Technical Summary: Digital Media Art Consumption and Symbolic Domestication Among Chinese Otaku Using a Multimodal Machine Learning Approach

Problem Statement
The Chinese Otaku community, now numbering in the hundreds of millions, represents a core consumer demographic driving the ACGN (Animation, Comics, Games, Novels) industry. However, existing theoretical frameworks, particularly those derived from Japanese Otaku culture (e.g., Hiroki Azuma's "database consumption" and "animalization"), fail to adequately explain the unique cultural logic of Chinese consumers. In China, the "animalization" of Otaku appears to converge "fundamentalism" (grand narratives) with the "database" (fragmented symbols), creating a distinct "fundamentalist database." Furthermore, the interplay between commercial capital and subcultural groups involves a complex process of "symbolic domestication" and "reverse domestication" within the digital media art field. There is a lack of empirical, quantitative research that integrates multimodal data (text and images) to identify the specific drivers of consumption intention and affective identification among Chinese Otaku, particularly regarding the roles of "System Narrative," "Guochao" (National Tide) integration, and community co-creation.

Methodology
This study employs a multimodal machine learning approach to analyze a large-scale dataset harvested from Bilibili, Xiaohongshu, and Weibo Super Topics between January 2024 and March 2026.

Data Collection: The final corpus consists of 81,457 valid textual comments and 123,692 associated images covering 127 core ACGN works/IPs. The sample is stratified by generation, with 71.5% Generation Z and 28.5% Millennials.
Labeling System: A three-tier labeling system was constructed based on theory-driven and data-driven principles. Key second-tier features include "System Narrative," "Moe Element Consumption," "Guochao Integration," "Community Co-creation," "Media Identification Intensity," and "Indigenous Cultural Identification."
Multimodal Processing:
- Text: Processed using Jieba segmentation with a custom ACGN lexicon, noise filtering, and BosonNLP for sentiment analysis.
- Images: Standardized to 640×640 pixels, augmented via rotation/flipping, and deduplicated using perceptual hashing.
- Label Mapping: Text was mapped using SentenceTransformer (cosine similarity threshold $\tau=0.6$ ), and images were mapped using the CLIP model (ViT-B/32, $\tau=0.65$ ).
- Fusion: A weighted fusion strategy (0.6 for text, 0.4 for image) was applied to create a six-dimensional feature vector for each sample.
Machine Learning Model: An XGBoost ensemble model was trained to predict two continuous targets: Consumption Intention and Sentiment Value. Hyperparameters were optimized via grid search and 5-fold cross-validation.
Interpretability: SHapley Additive exPlanations (SHAP) were used to quantify feature importance and analyze marginal effects, addressing the "black box" nature of XGBoost.
Statistical Analysis: Pearson correlation, mediation analysis (PROCESS Model 4), and moderation analysis (PROCESS Model 1) were conducted to validate theoretical hypotheses.

Key Results

Multimodal Necessity: Multimodal fusion significantly enhanced feature coverage, with an average increase of 186.5%. "System Narrative" and "Guochao Integration" saw coverage jumps of 285.4% and 192.6%, respectively, demonstrating that visual data is critical for capturing digital art consumption characteristics.
System Narrative Dominance: Contrary to the "Moe element database" model, "System Narrative" (cultivation systems, worldbuilding rules) is a fundamental driver. While "Moe Element Consumption" had a slightly higher global SHAP value for sentiment, "System Narrative" showed a stronger standardized regression coefficient ( $\beta=0.287$ ) on consumption intention and a steeper marginal effect for high-intent cohorts.
Mediation of Guochao: "Indigenous Cultural Identification" partially mediates the relationship between "Guochao Integration" and "Consumption Intention," accounting for 39.8% of the total effect. This confirms that cultural confidence acts as a psychological bridge transforming aesthetic appreciation into purchase behavior.
Moderation of Co-creation: "Community Co-creation Intensity" positively moderates the effect of Guochao integration on consumption. High levels of community participation amplify the driving effect of cultural elements.
Generational Divergence:
- Generation Z: Driven by "Moe Element Consumption" and "Community Co-creation" (visual-community model). They show higher sensitivity to Guochao elements and Indigenous Cultural Identification.
- Millennials: Driven by "System Narrative" and "Media Identification Intensity" (narrative-identification model), valuing narrative coherence and deep immersion.

Key Contributions

Theoretical Innovation: The study constructs an indigenized "Context—Field—Identification—Behavior" model, challenging the direct transposition of Japanese Otaku theories. It proposes the concept of a Chinese "fundamentalist database" where grand narratives and fragmented symbols coexist.
Methodological Innovation: It demonstrates the efficacy of integrating XGBoost with SHAP on multimodal (text + image) data to quantify subcultural consumption patterns, moving beyond traditional linear regression to capture non-linear feature interactions.
Empirical Insight: The research provides quantitative evidence for the "System Narrative Dominance" in Chinese Otaku culture and elucidates the bidirectional mechanism of "symbolic domestication" (capital guiding users vs. users re-signifying content).

Significance and Claims
The paper claims to offer novel empirical and theoretical insights into subcultural consumption within the digital media art field. By validating the "System Narrative Dominance," it suggests that Chinese Otaku consumption is characterized by a pursuit of systemic rationality and order, distinct from the fragmented consumption of postmodern Japanese contexts. The study asserts that "Guochao Integration" is not merely a visual trend but a mechanism for activating cultural identity, which is further amplified by participatory culture.

Practically, the authors claim these findings provide data-driven strategies for the ACGN industry:

Content Strategy: Shift from simple "Moe accumulation" to constructing coherent "System" worldviews and organically integrating cultural symbols (e.g., Taoist philosophy as cultivation logic).
User Operations: Adopt differentiated strategies for Generation Z (visual/interactive focus) and Millennials (narrative depth focus) and leverage "Community Co-creation" as an amplifier for IP value.

The study acknowledges limitations, including the exclusion of private community data (e.g., QQ groups), the cross-sectional nature of the data preventing longitudinal analysis of "symbolic domestication," and the need for future cross-cultural comparisons (e.g., with South Korea) to further validate the distinctiveness of the "System Narrative" phenomenon.