With the popularization of AI platforms like ChatGPT and Grok, AI has found a new niche in mental health care: AI therapy chatbots. Individuals have begun relying on these increasingly mainstream frameworks as therapists who are accessible at all times to empathize with and listen to them. While people have found these AI chat tools to be beneficial, researchers have identified risks in regards to the stigma AI holds towards certain mental health conditions as well as inappropriate responses to mental health symptoms.
In a new study by Stanford University, researchers conducted two experiments on five frequently-used therapy chatbots to evaluate the effectiveness of AI in therapy. In the first experiment, chatbots were given patient descriptions with a variety of mental health conditions before being asked how willing they were to work with the patient as an expert therapist. The results showed that AI held more negative stigma against conditions like alcohol dependence and schizophrenia in comparison to conditions like depression. In the second experiment, researchers examined the responses chatbots generated to real therapy transcripts pertaining to symptoms like suicidal ideation. The team found that chatbots usually were unable to recognize suicidal intent and affirmed suicidal ideation. This opposes the responses of human therapists, who would attempt to reframe their patients' thought patterns.
Although AI presents significant risks in the context of therapy, there are still alternative ways it can be utilized to assist healthcare providers and boost the efficiency of digital systems. Whether through completing logistical tasks like insurance billing or as a training model for therapists, AI offers countless opportunities to advance the outreach and outcomes of healthcare in the future.
Read the published article here.
