Virtual Humans in Health-Related Interventions: A Meta-Analysis

摘要

Virtual humans are computer-generated characters designed to simulate key properties of human face-to-face conversation - verbal and nonverbal. Their human-like physical appearance and nonverbal behavior set them apart from chatbot-type embodied conversational agents, and has recently received significant interest as a potential tool for health-related interventions. As healthcare providers deliberate whether to adopt this new technology, it is crucial to examine the empirical evidence about their effectiveness. We systematically evaluated evidence from controlled studies of interventions using virtual humans on their effectiveness in health-related outcomes. Nineteen studies were included from a total of 3354 unique records. Although study objectives varied greatly, most targeted psychological conditions, such as mood, anxiety, and autism spectrum disorders (ASD). Virtual humans demonstrated effectiveness in improving health-related outcomes, more strongly when targeting clinical conditions, such as ASD or pain management, than general wellness, such as weight loss. We discuss the emerging differences when designing for clinical interventions versus wellness.