Word Embeddings Track Social Group Changes Across 70 Years in China
This study presents the first large-scale computational analysis of Chinese state-controlled media from 1950 to 2019, revealing that while certain social stereotypes remain stable, official linguistic representations of gender and economic classes undergo dramatic shifts that track China's revolutionary social transformations, thereby offering a crucial non-Western perspective on how language encodes societal change.
Original paper licensed under CC BY 4.0 (http://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 language as a giant, living mirror. Usually, we think of this mirror just reflecting what society looks like. But this paper argues that the mirror also acts like an architect: the way we talk about people actually helps build and shape how society sees them.
The researchers wanted to see how this "mirror" changed over 70 years in China (from 1950 to 2019). They didn't just read newspapers; they used a special computer tool called Word Embeddings. Think of this tool as a giant, invisible map where words are like cities. If two words are often used together (like "doctor" and "smart"), their cities are built close to each other. If they are rarely used together, they are far apart.
By looking at this map over seven decades, the team tracked how the "cities" of different social groups moved, shrank, or changed their neighborhoods. Here is what they found, broken down simply:
1. The Two Types of Mirrors
The study looked at two main types of changes in how people were described in official Chinese newspapers (specifically The People's Daily, which acts as the government's official voice).
The "Sticky" Groups (Things that didn't change much)
Some social groups were like heavy furniture that was hard to move, even when the room was being completely renovated. No matter what political storms happened, the way these groups were described stayed surprisingly consistent:
- Young vs. Old: Young people were always seen as energetic and capable, while older people were consistently described with negative traits like being confused or rigid.
- Thin vs. Fat: Thin people were linked to positive words like "strong" and "dignified," while fat people were linked to negative words like "bad" or "stupid."
- Ethnic Minorities vs. Han Chinese: The majority group (Han Chinese) was often described with neutral or slightly negative practical traits, while ethnic minorities were consistently painted with a "positive" but simplistic lens—always "happy," "simple," and "pure."
The "Shapeshifters" (Things that flipped upside down)
Other groups were like acrobats, flipping their descriptions dramatically depending on the political era:
- Gender: In the 1950s and 60s, women were described as innocent and gentle. But in the 1970s (during the Cultural Revolution), the description suddenly flipped to be harsh and negative (words like "evil" or "shameless"). By the 1990s, they were positive again, but by the 2010s, the balance shifted again, with men taking on more positive traits than women for the first time in decades.
- Rich vs. Poor: Before 1978, being "poor" was a badge of honor (linked to "hardworking" and "selfless"), while being "rich" was a crime (linked to "greedy" and "dictatorial"). But after the economic reforms in 1978, the description of the poor started to turn negative, while the rich became slightly less hated, though still not fully "good."
2. The "Earthquake" Years
The researchers found that language usually changes slowly, like a river carving a canyon over centuries. However, they discovered a specific period (1966–1976, the Cultural Revolution) where the river suddenly turned into a tsunami.
During these ten years, the connection between words and meanings broke down. The computer map showed that the way people were described changed so fast and so violently that the usual "stability" of language disappeared. It was a time when the government's voice completely rewrote the dictionary of social value overnight.
3. How China Differs from the West
The paper compares these findings to similar studies done in the US.
- In the US: The "Rich" are usually described positively (successful, smart), and minorities are often described negatively.
- In China (Official Media): The "Rich" are described negatively (corrupt, greedy) because of socialist ideology, even after the country opened up its markets. Meanwhile, minorities are described with "positive" stereotypes (simple, happy), which the authors call "positive othering"—it sounds nice, but it still treats them as childlike or primitive compared to the mainstream.
The Big Takeaway
This study shows that language isn't just a passive recording of history; it's an active tool used by the state to rebuild society.
- Some deep-seated biases (like judging people by their age or weight) are so strong that even massive political revolutions couldn't erase them.
- Other biases (like how we view gender or money) are fragile and can be completely rewritten by political power.
The authors conclude that to truly understand how societies work, we can't just look at the West. We need to study places like China, where the "mirror" was deliberately smashed and remade multiple times in a single lifetime, showing us exactly how language can be used to construct a new reality.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.