Tag
This paper uses GPT-4.1 to annotate 9,000 support conversations and decompose customer satisfaction into component axes, validating the annotations against self-reported ratings and revealing lower satisfaction in full-census data compared to survey responses.
This paper investigates whether topic sentiment causally affects perceived political ideology in news articles, comparing human annotations from AllSides with those from LLMs including GPT-4o-mini and Llama-3.3-70B. It finds that fine-tuned GPT-4o-mini exhibits a spurious sentiment-ideology coupling not present in human judgments, highlighting risks of using LLM annotations as proxies in causal analyses.