Fine-tuning Content Moderation for Fictional Storytelling

One of the benefits of Subtxt over something like ChatGPT or Claude is the ability to fine-tune or adjust the level of Content Moderation that goes on in the background. This came up today in the Workshop where a student’s story was flagged as being too violent. Upon investigation, the AI was a bit nervous as to the source material (it was a story about some nasty vampires), but I found it perfectly acceptable.

In response, I adjusted the filter for violent material in Subtxt and Subtxt Muse to allow for this kind of story. I also went ahead and updated the Guide to explain the reasoning behind this fix.

James R. Hull @jhull