ChatGPT censors books with sexual content in Iowa schools. The educational staff of the Mason City School District proceeded to select the objectionable books according to a new law passed in May using the famous chatbot.
In the United States, it is not only in New York that ChatGPT is used in schools. In Iowa, it is not the students who employ him, but the administrative staff. But in this case, ChatGPT does not prevent the correction of exam copies: the OpenAI chatbot is used to choose which books to censor from school district libraries.
Governed by Republican Party conservatives, Iowa recently passed laws against books with sexual content in schools. In Mason City, school libraries must now comply, removing all books that meet strict criteria. To determine whether books should be censored, staff relied heavily on ChatGPT.
A prompt ChatGPT to detect books to censor
To find books with objectionable content, administrators first compiled an initial list of suspicious books, using sources such as the press and Wikipedia, and then crossed it with titles in library collections. They then seized the question “Does this book contain a description or representation of a sexual act?” in ChatGPT for fifty suspicious books.
When asked by Popular Science, the school district’s response leaves no doubt: “If the answer is yes, the book is taken out of circulation and stored,” says Bridgette Exman, deputy superintendent of the Mason City school program and education. Among the books that did not pass the ChatGPT filter is Margaret Atwood’s bestseller The Handmaid’s Tale, which recently received a critically acclaimed television adaptation.
The superintendent told Popular Science that reading every book on the list was an insurmountable task. The new school year is approaching, yet education staff say the state has not provided clear guidance on how to implement changes to the curriculum.
Mason City administrators claim to have submitted ChatGPT’s list to a librarian for confirmation. Indeed, the answers generated by artificial intelligence are sometimes contradictory depending on the prompt used. Thus, The Verge reports testing the 19 books banned by the district in ChatGPT, asking if they have erotic passages. The software would then have indicated that several of them did not contain any.
Post a Comment