Use of ChatGPT | Key findings | References |
---|---|---|
Teaching in social psychiatry | • ChatGTP was found to have the following functions: ▪ Acting as an information provider ▪ Used as a tool for debate and discussions ▪ Used in content creation for course materials ▪ Generate hypothetical case vignettes in the area of social psychiatry | [13] |
Clinical and ethical reasoning, diagnosis, treatment and prognosis using psychiatry case vignettes | • The study investigated ChatGPT’s ability in clinical and ethical reasoning, diagnosis, treatment and prognosis using 100 psychiatry case vignettes, with the following performance: ▪ Scored grade A for 61 cases ▪ Scored grade B for 31 cases ▪ Scored grade C for 8 cases ▪ No responses were graded  "D" | [14] |
Generation of psychodynamic formulation for psychiatric cases | • ChatGPT could generate psychodynamic formulations that were rated appropriate by psychiatrists • The formulations were further improved by additional instructions such as self-psychology, ego-psychology and object relations | [15] |
Answering questions about clinical psychiatry | • The study compared answers from participants (psychiatrists and psychiatry residents) regarding clinical psychiatry questions using information from ChatGPT and other sources • Participants who used ChatGPT outperformed those who used other sources of information • ChatGPT performed well with an overall score of 8 out of 10 for accuracy, nuance and completeness | [16] |
Diagnosis and treatment recommendation | • A hypothetical psychiatric case was presented to ChatGPT • The performance of ChatGPT was as follows: ▪ Correctly gave the diagnosis of treatment-resistant schizophrenia ▪ Provided relevant tests to rule of other causes of psychosis ▪ Proposed treatment plans consistent with current standards of care, including medications ▪ Provided a comprehensive list of potential side effects of the proposed medications | [17] |
Use of ChatGPT to train a machine learning (ML) model for identification of a psychiatric disorder | • The study used ChatGPT to train an ML model for identification of post-traumatic stress disorder (PTSD) after childbirth • The ML model was able to identify childbirth-related post-traumatic stress disorder (PTSD) based on maternal childbirth narrative data | [18] |