Using AI-Based Virtual Simulated Patients for Training in Psychopathological Interviewing: Cross-Sectional...

Area of Science:

• Medical Education
• Artificial Intelligence in Healthcare
• Psychology Training

Background:

• Generative artificial intelligence (GAI) virtual simulated patients (VSPs) show potential for training clinical interviewing skills.
• Little is known about how system and user variables affect student perceptions of GAI-driven VSP interactions.

Purpose of the Study:

• To investigate psychology students' perceptions of GAI-driven VSPs.
• To examine the influence of demographic factors, system parameters, and interaction characteristics on these perceptions.

Main Methods:

• 1832 interactions between 156 psychology students and 13 GAI VSPs with varied temperature settings (0.1, 0.5, 0.9).
• Data collected included student demographics, interview length, connectivity failures, VSP used, and temperature setting.
• Student ratings, comments, and perceived diagnostic ability were recorded; sentiment analysis assessed clinical realism.

Main Results:

• Female students rated VSPs higher (9.25/10) than male students (8.94/10).
• Higher ratings correlated with fewer connectivity failures and higher GAI temperature settings (0.9 vs. 0.1).
• 94% of students perceived improved diagnostic ability, but objective scores did not significantly differ from a non-VSP cohort.

Conclusions:

• GAI-driven VSPs are positively received by psychology students, with gender and system variables influencing evaluations.
• While students felt training improved diagnostic skills, objective outcomes showed no significant gains.
• Further research is needed due to limitations in randomization and generalization.