>>2304I don't think it's really a lack of nuance, just that LLMs are terrible at seeing
statistical significance. It'll probably just parrot back the p-values and confidence intervals that are already in the report. The text generation might be low effort, but it's not inventing new logic. It's just going to give us a fancy way to read the same boring summary we're already looking at. I expect it to be mostly redundant. Basically just a different font for the same bad news.