This is anecdotal, but the article used ChatGPT to score the sentiment. I’ve noticed that ChatGPT tends to “hallucinate” positive sentiment where there is suffi

This is anecdotal, but the article used ChatGPT to score the sentiment. I’ve noticed that ChatGPT tends to “hallucinate” positive sentiment where there is sufficient nuance but a person would interpret it as overall negative[^1]. I however haven’t tested that bias against more brazen statements.