17 sources behind this chapter. Each one carries a note on what it is doing in the argument.
CACM article used for the chapter’s practical example of AI output feeding later AI workflows and flattening useful differences.
Open this source Read the original
Donska, Tanya. ‘When AI Tools Train on AI Output: Model Collapse in Daily Workflows.’ Communications of the ACM, Blog@CACM, 6 February 2026.
Alex Williams’s essay brings model collapse out of the laboratory and onto the open web: synthetic material is already entering the pool from which later systems learn, sanding down rare and useful differences.
Open this source Read the original
Williams, Alex. 'Model Collapse Is Already Happening, We Just Pretend It Isn’t.' Communications of the ACM, Blog@CACM, 25 March 2026. https://cacm.acm.org/blogcacm/model-collapse-is-already-happening-we-just-pretend-it-isnt/.
The peer-reviewed foundation for the model-collapse discussion. Repeated training on generated data can erase the tails of the original distribution until the model retains a thinner version of reality.
Open this source Read the original
Shumailov, I., Shumaylov, Z., Zhao, Y., Papernot, N., Anderson, R., and Gal, Y. 'AI Models Collapse When Trained on Recursively Generated Data.' Nature 631 (July 2024): 755-759. doi.org/10.1038/s41586-024-07566-y
The earlier research version of the model-collapse finding, showing how recursive training makes models forget low-probability parts of the original data.
Open this source Read the original
Shumailov, I., et al. 'The Curse of Recursion: Training on Generated Data Makes Models Forget.' arXiv:2305.17493, 2023.
Directly cited in Chapters 7, 8 and 16. Its broken junior-development pipeline informed Chapter 24 during development, but that prose block was removed; retained under Other Sources as supporting background.
Starkey, Richard. The 4th Industrial Revolution: Artificial Intelligence in Film Post Production and the Impact on the Creative Economy. MFA research report, AFDA Cape Town, 2020.
This is the source for Cory Doctorow’s term for the familiar platform cycle in which a useful service is steadily made worse for users, then businesses, in pursuit of extraction.
Open this source Read the original
Doctorow, Cory. "The 'Enshittification' of TikTok." Wired, 23 January 2023. (Term first used on his Pluralistic blog, late November 2022.) https://www.wired.com/story/tiktok-platforms-cory-doctorow/
Used to show how quickly Doctorow’s wonderfully ugly word escaped the tech world and became the American Dialect Society’s 2023 Word of the Year.
Open this source Read the original
American Dialect Society. 2023 Word of the Year Is "Enshittification." 5 January 2024. https://americandialect.org/2023-word-of-the-year-is-enshittification/
Doctorow’s full-length account of how useful platforms are degraded by extraction. It gives the flattening chapter a wider economic pattern beyond AI-generated prose.
Open this source Read the original
Doctorow, Cory. Enshittification: Why Everything Suddenly Got Worse and What to Do About It. First ed. New York: MCD, Farrar, Straus and Giroux, 2025. London: Verso Books, 2025.