📄 social_science

Structural Anxiety and Gender Heterogeneity in Public Fertility Discourse: Topic and Sentiment Analysis of Weibo Data Using the T-SENCE Framework

Using the T-SENCE framework to analyze over 55,000 Weibo posts, this study reveals that public fertility discourse is dominated by structural anxieties and exhibits profound gender heterogeneity, with women expressing internalized emotional tensions and men adopting externalized rational perspectives, thereby highlighting the urgent need to address asymmetric structural pressures to foster a fertility-friendly society.

Original authors: Wenjie Tang, Jianhao Ma, Yujia Hu, Xia Wang, Jia Shi, Yuehua Yang, Ruochen Huang

Published 2026-09-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wenjie Tang, Jianhao Ma, Yujia Hu, Xia Wang, Jia Shi, Yuehua Yang, Ruochen Huang

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

In the modern world, the decision to have children has shifted from a simple, expected step in life to a complex calculation weighed against personal freedom, financial stability, and career goals. This change is part of a broader global trend where people are prioritizing their own lives and values over traditional family structures, a phenomenon sociologists call the "Second Demographic Transition." At the same time, a concept known as the "incomplete gender revolution" suggests that while women have gained more power in the workplace, the heavy lifting of domestic life and the risks of pregnancy still fall disproportionately on them. When these two forces meet, they create a unique kind of pressure. People are not just deciding whether to have a family; they are navigating a landscape where the costs of doing so feel incredibly high, and the emotional toll is unevenly shared between men and women. Understanding how people talk about these pressures in public spaces offers a window into the real fears and hopes shaping the future of families.

To uncover these hidden dynamics, researchers turned to the digital public square of Weibo, China's largest microblogging platform. They collected and analyzed over 55,000 individual posts written over the course of a single year, from May 2024 to April 2025. Instead of relying on simple surveys or basic computer programs that might miss the nuance of human language, the team used a sophisticated new approach called T-SENCE. This method combines two powerful tools: one that maps out the main topics people are discussing, and another that uses advanced artificial intelligence to understand the deep emotional context of what is being said, including sarcasm, fear, or genuine hope. By feeding this massive amount of text into the system, the researchers could see not just what people were talking about, but how they felt about it, and whether men and women were experiencing the same reality.

The analysis revealed that public conversation about marriage and children has fractured into nine distinct themes, moving far beyond the simple question of "when to have a baby." The most dominant topic, making up nearly one-fifth of all discussions, was "Social Observations and Emotional Aspirations." This category captures the way people react to the lives of their peers, such as seeing a friend get married or hearing about a happy family. It shows that individual decisions are deeply influenced by watching the lives of others. The next most common themes were "Family of Origin and Reproductive Responsibility," which deals with the pressure from parents to continue the family line, and "Personal Experience of Pregnancy and Childbirth," which focuses on the physical and emotional realities of having a baby. Together, these topics paint a picture of a society where the decision to reproduce is a constant negotiation between personal desires, family expectations, and the harsh realities of daily life.

When the researchers looked at the emotional tone of these conversations, they found a landscape dominated by worry. More than half of all the posts, specifically 50.31 percent, carried a negative sentiment. This negativity was not a general sadness about life, but a focused response to specific structural pressures. The most anxious discussions centered on the heavy costs of raising a child, the unfairness of gender roles, and the physical pain and risks associated with pregnancy. In contrast, only about 23 percent of the posts were positive. The few areas where hope appeared were in the stories of others' happiness, where people found temporary comfort in the idealized images of weddings and family life shared online. This suggests that while the digital space is filled with fear and stress, it also serves as a place where people cling to the dream of a better future, even as they struggle with the present.

Perhaps the most striking discovery was how differently men and women experienced this digital conversation. Women adopted what the researchers call an "internalized perspective," treating the decision to have a child as a deeply personal, bodily experience. They spoke more frequently about the physical realities of pregnancy, the pain of childbirth, and the fear of losing their own identity or career opportunities. Their emotions were intense and polarized; they expressed the highest levels of fear regarding the sacrifices required, yet they also held the strongest hopes for a romantic, ideal family life. This created a "dual-high" tension, where deep anxiety coexisted with a powerful yearning for happiness.

Men, on the other hand, viewed the issue through an "externalized perspective." Their discussions were less about personal bodily experiences and more about social responsibilities, such as providing for a family or meeting the expectations of their own parents. They were more likely to discuss the economic and societal structures surrounding child-rearing rather than the intimate details of pregnancy. Emotionally, men acted as a "rational ballast" in the conversation. They expressed significantly more neutral sentiment than women, often sounding like observers or analysts rather than direct participants in the emotional turmoil. While women felt the weight of the decision in their bodies and their daily lives, men tended to approach the topic with a more detached, factual tone.

This divide highlights a profound disconnect in how the two genders navigate the path to parenthood. The study suggests that the digital world amplifies these differences, creating a space where women feel the full, raw intensity of the risks and costs, while men remain somewhat insulated by their position in the current social structure. The research does not claim that one group is right or wrong, but rather that they are living in two different emotional realities within the same society. The findings confirm that the shift away from traditional family planning is driven by a genuine, structural anxiety about the costs of raising a child, and that this anxiety is felt much more acutely by women. To build a society where people feel comfortable having children, the researchers argue, we must address these uneven burdens and the deep structural pressures that currently dominate the public conversation.

Technical Summary: Structural Anxiety and Gender Heterogeneity in Public Fertility Discourse

Problem Statement

Global and Chinese fertility rates are declining amidst rapid demographic shifts, driven by a transition from normative societal expectations to individualized value assessments. While the "Second Demographic Transition" (SDT) and the "incomplete gender revolution" provide theoretical frameworks for these changes, existing empirical studies face significant limitations. Current research often relies on small-scale self-reported surveys that fail to capture spontaneous digital discourse, or utilizes traditional computational models (e.g., standard sentiment dictionaries) that struggle with the irony, metaphors, and context-dependent contradictions prevalent in Chinese social media. There is a critical gap in understanding the macro-thematic landscape and the micro-level, highly polarized emotional complexities—specifically gendered nuances—of public fertility discussions.

Methodology: The T-SENCE Framework

To address these gaps, the authors propose the T-SENCE (Topic-Sentiment Extraction analysis with Natural language Contextual Evaluation) framework. This integrated pipeline analyzes a robust corpus of 55,032 valid, individual-level Weibo posts collected over a one-year period (May 1, 2024, to April 30, 2025). The methodology consists of four synergistic modules:

Data Acquisition and Corpus Construction: A custom crawler retrieved posts using 22 high-relevance keywords (e.g., "childbearing," "plan to marry"). Data underwent deduplication and text standardization (removing URLs, emojis, and special characters) to ensure a clean, individual-level dataset.
Latent Topic Extraction (Macro-Mapping): The study employed Latent Dirichlet Allocation (LDA) to map the structural breadth of the discourse.
- Preprocessing: Text was segmented using the Jieba library, with stop words filtered via a hybrid of the HIT stop-word list and a custom domain-specific dictionary.
- Optimization: The optimal number of topics ( $K$ ) was determined by minimizing Perplexity and maximizing the Coherence Score (UMass), resulting in $K=9$ .
- Human-AI Collaboration: An instruction-fine-tuned Large Language Model (Qwen2.5-32B) generated preliminary semantic labels for the 9 extracted topics, which were then rigorously reviewed and calibrated by a panel of public health and sociology experts to ensure contextual validity.
Natural Language Contextual Sentiment Evaluation (Micro-Perception): To overcome the limitations of traditional lexicon-based tools, the study utilized Qwen2.5-32B for context-aware ternary sentiment classification (Positive, Neutral, Negative).
- Approach: The model was guided via few-shot learning with precise instruction prompts to handle complex sociological contexts, internet slang, and irony.
- Output: Each post was assigned a sentiment label, moving beyond simple polarity to capture deep semantic understanding.
Multidimensional Integration and Demographic Analysis: The framework aligned LDA topic probability vectors with LLM sentiment labels and user gender attributes (parsed from metadata).
- Analysis: A cross-tabulation analysis calculated conditional probability distributions of emotions under specific topics.
- Stratification: The joint matrix was bifurcated into male and female cohorts to compare topic engagement and asymmetric emotional resonance.

Key Results

1. Thematic Composition

The analysis identified nine core themes dominating public discourse:

Social Observations and Emotional Aspirations (20.60%): The dominant theme, driven by peer comparison and social referencing.
Family of Origin and Reproductive Responsibility (16.97%): Reflecting intergenerational pressures.
Personal Experience of Pregnancy and Childbirth (14.79%): Focusing on individual, embodied narratives.
Other themes included Marital Challenges, Life Meaning, Gender Dynamics, Romantic Beliefs, Career Development, and Post-marital Roles.

2. Sentiment Landscape

The overall sentiment exhibited a severe negative deviation:

Negative: 50.31%
Neutral: 26.14%
Positive: 23.55%
Negative sentiment was heavily concentrated in topics related to structural pressures and individual costs. The highest negative proportions were found in "Family of Origin" (60.59%), "Gender Dynamics" (60.09%), and "Personal Experience of Pregnancy" (58.65%). Conversely, "Social Observations" was the only category where positive sentiment (35.40%) nearly matched negative sentiment (38.78%), driven by idealized scenarios of peers' happiness.

3. Gender Heterogeneity

Profound differences emerged between male and female users:

Female Users ("Internalized Perspective"):
- Focus: Heavily engaged in "Personal Experience of Pregnancy and Childbirth" (15.64% vs. 11.22% for males) and "Life Meaning."
- Emotion: Exhibited a "dual-high" emotional tension. They expressed significantly higher negative sentiment regarding structural inequalities (e.g., 62.88% negative in "Family of Origin") while simultaneously projecting stronger positive emotions toward idealized scenarios (36.35% positive in "Social Observations").
Male Users ("Externalized Perspective"):
- Focus: Dominated by "Family of Origin and Reproductive Responsibility" (21.87% vs. 15.81% for females) and macro-structural topics like "Gender Dynamics."
- Emotion: Acted as a "rational ballast," displaying a higher prevalence of neutral sentiment (e.g., 34.39% neutral in "Family of Origin" vs. 25.32% for females). Their discourse was more observational and less emotionally charged regarding personal sacrifice.

Significance and Claims

The paper claims to provide robust empirical evidence validating the "Second Demographic Transition" and the "incomplete gender revolution" within the digital era.

Theoretical Validation: The findings confirm that public fertility discourse has shifted from viewing childbearing as a standardized life task to a refined, risk-averse calculation involving structural pressures, workplace costs, and gender rights.
Mechanism of Anxiety: The study identifies "social referencing" on social media as a critical amplifier. While digital spaces transmit contradictory extremes (fear of pain vs. joy of motherhood), they also create a "misaligned resonance" where women face a "dual-high" tension of anxiety and yearning, while men maintain a detached, rational stance.
Policy Implication: The authors conclude that building a genuinely fertility-friendly society requires fundamentally alleviating the asymmetric structural anxieties and profound gendered costs (such as the "motherhood penalty") that currently dominate public consciousness. The data suggests that current policies may not be addressing the specific, embodied, and internalized fears of women, which are distinct from the externalized, responsibility-focused concerns of men.

The study acknowledges limitations regarding platform representation bias (Weibo skews young), algorithmic amplification of extreme sentiments, and the lack of multimodal data (images/videos), suggesting these as areas for future infodemiology research.