Fine-Tuned LLMs are “Time Capsules” for Tracking Societal Bias Through Books
Sangmitra Madhusudan, Robert Morabito, Skye Reid, Nikta Gohari Sadr, and Ali Emami · 2025
Read Fine-Tuned LLMs are “Time Capsules” for Tracking Societal Bias Through Books by Sangmitra Madhusudan, Robert Morabito, Skye Reid, Nikta Gohari Sadr, and Ali Emami (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-05331.txt.
Description
Novels carry the biases of their time, and language models trained on them may soak those biases up. Sangmitra Madhusudan and co-authors turn that risk into a measuring tool in a 2025 study of fiction across seventy years.
The team built a corpus of 593 novels from the 1950s to the 2010s and fine-tuned models on each decade's books. Targeted prompts then revealed how each model's views on gender, sexuality, race, and religion shifted by decade. Portrayals of women in leadership rose over time, and same-sex references grew sharply after the 1980s. Negative portrayals of Islam jumped in the 2000s. Control tests showed the books, not the base models, drove the patterns.
Catalog record
| CREATOR | Sangmitra Madhusudan, Robert Morabito, Skye Reid, Nikta Gohari Sadr, and Ali Emami |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-05331