One of the most mind-blowing ideas I’ve encountered lately is that computers can help us measure cultural meaning. Specifically, researchers use a method called word embeddings (a type of natural language processing that places words in a mathematical space based on how often they appear near other words). Words with similar meanings tend to cluster together. But here’s the fascinating part: this method reveals our cultural biases, both past and present.

A 2018 study titled The Geometry of Culture analyzed massive text corpora, such as books, newspapers, and academic journals, spanning more than 100 years in the U.S. and U.K. The researchers measured how semantic relationships between concepts like “man-woman” or “rich-poor” changed over time.

The key insight? Culture exists in the space between words. By analyzing how “doctor” relates to “man” versus “woman” in different decades, you can quantify shifting gender norms. These semantic shifts reflect real social changes -even before they show up in law or policy.

This method has also been used to study racial bias, political ideology, and even the evolution of religious language. For example, how closely “God” is associated with “king” versus “parent” across time reveals changes in theological imagination.

Of course, word embeddings have limitations. They can reproduce harmful stereotypes if used uncritically. But when paired with critical theory and context, they become powerful tools for cultural anthropology and gender studies.

Language isn’t neutral, it’s a mirror of our shared assumptions. And with these tools, we can now watch that mirror shift in real time.