Tag
This paper evaluates the safety risks of therapy bots for Generation Alpha, finding that large language models struggle to calibrate clinical risk correctly, leading to a significant vocabulary-comprehension gap that could result in missed crises.