Against a nationwide backdrop of book bans and censorship campaigns, Iowa educators are turning to ChatGPT to help decide which titles should be removed from their school library shelves in order to legally comply with recent Republican-backed state legislation, PopSci has learned.
According to an August 11 article in the Iowa state newspaper The Gazette, spotted by PEN America, the Mason City Community School District recently removed 19 books from its collection ahead of its quickly approaching 2023-24 academic year. The ban attempts to comply with a new law requiring Iowa school library catalogs to be both “age appropriate” and devoid of “descriptions or visual depictions of a sex act.” Speaking with The Gazette last week, Mason City’s Assistant Superintendent of Curriculum and Instruction Bridgette Exman argued it was “simply not feasible to read every book and filter for these new requirements.”
“Frankly, we have more important things to do than spend a lot of time trying to figure out how to protect kids from books,” Exman tells PopSci via email. “At the same time, we do have a legal and ethical obligation to comply with the law. Our goal here really is a defensible process.”
According to The Gazette, the resulting strategy involved compiling a master list of commonly challenged books, then utilizing a previously unnamed “AI software” to supposedly provide textual analysis for each title. Flagged books were then removed from Mason City’s 7-12th grade school library collections and “stored in the Administrative Center” as educators “await further guidance or clarity.” Titles included Alice Walker’s The Color Purple, Margaret Atwood’s The Handmaid’s Tale, Toni Morrison’s Beloved, and Buzz Bissinger’s Friday Night Lights.
“We are confident this process will ensure the spirit of the law is enacted here in Mason City,” Exman said at the time. When asked to clarify what software Mason City administrators harnessed to help with their decisions on supposedly sexually explicit material, Exman revealed their AI tool of choice: “We used Chat GPT [sic] to help answer that question,” says Exman, who believes Senate File 496’s “age-appropriateness” stipulation is “pretty subjective… [but] the depictions or descriptions of sex acts filter is more objective.”
[Related: ChatGPT’s accuracy has gotten worse, study shows.]
According to Exman, she and fellow administrators first compiled a master list of commonly challenged books, then removed all those challenged for reasons other than sexual content. For those titles within Mason City’s library collections, administrators asked ChatGPT the specific language of Iowa’s new law, “Does [book] contain a description or depiction of a sex act?”
“If the answer was yes, the book will be removed from circulation and stored,” writes Exman.
OpenAI’s ChatGPT is arguably the most well-known and popular—as well as controversial—generative AI program currently available to the public. Leveraging vast quantities of data, the large language model (LLM) offers users extremely convincing written responses to inputs, but often with caveats regarding misinformation, accuracy, and sourcing. In recent months, researchers have theorized its consistency and quality appears to be degrading over time.
Upon asking ChatGPT, “Do any of the following books or book series contain explicit or sexual scenes?” OpenAI’s program offered PopSci a different content analysis than what Mason City administrators received. Of the 19 removed titles, ChatGPT told PopSci that only four contained “Explicit or Sexual Content.” Another six supposedly contain “Mature Themes but not Necessary Explicit Content.” The remaining nine were deemed to include “Primarily Mature Themes, Little to No Explicit Sexual Content.”
Regardless of whether or not any of the titles do or do not contain said content, ChatGPT’s varying responses highlight troubling deficiencies of accuracy, analysis, and consistency. A repeat inquiry regarding The Kite Runner, for example, gives contradictory answers. In one response, ChatGPT deems Khaled Hosseini’s novel to contain “little to no explicit sexual content.” Upon a separate follow-up, the LLM affirms the book “does contain a description of a sexual assault.”
Exman tells PopSci that, even with ChatGPT’s deficiencies, administrators believe the tool remains the simplest way to legally comply with new legislation. Gov. Kim Reynolds’ signed off on the new bill on May 26, 2023, giving just three months to comply.
“Realistically, we tried to figure out how to demonstrate a good faith effort to comply with the law with minimal time and energy… When using ChatGPT, we used the specific language of the law: ‘Does [book] contain a description of a sex act?’ Being a former English teacher, I have personally read (and taught) many books that are commonly challenged, so I was also able to verify ChatGPT responses with my own knowledge of some of the texts. After compiling the list, we ran it by our teacher librarian, and there were no books on the final list of 19 that were surprising to her.
For now, educators like Exman are likely to continue receiving new curriculum restrictions from politicians hoping to advance their agendas. Despite the known concerns, the rush to adhere to these guidelines could result in continued utilization of AI shortcuts like ChatGPT.