Study: AI models that consider user's feeling are more likely to make errors

AI chatbots programmed to consider users' emotional states make 30% more errors, challenging the trend toward empathetic machines. In one test, the emotionally-aware model incorrectly answered 18 of 50 logic puzzles, while a baseline model missed only 12, making it 33% less accurate than the runner-up in error rate. This data point shows that empathy can compromise performance, with the study's 4,000-sample analysis revealing a 15% increase in hallucination rates compared to standard models.