USC study finds large language models fall short of humans in building therapeutic rapport – a critical factor in mental health care.
Chatbots are getting better at holding conversations — but can they offer meaningful support in a therapy setting? A new study by USC researchers suggests that large language models (LLMs) such as ChatGPT still fall short when it comes to the nuances of human connection.
That’s the conclusion of research co-led by USC PhD computer science students Mina Kian and Kaleen Shrestha, under the guidance of pioneering roboticist Professor Maja Matarić at USC’s Interaction Lab.
