Never say never: Exploring the effects of available knowledge on agent persuasiveness in controlled physiotherapy motivation dialogues

Researchers at an academic institution announced a new study on February 12, 2026, examining how the knowledge available to Generative Social Agents (GSAs) affects their persuasive capabilities in physiotherapy motivation dialogues, aiming to understand the balance between these AI systems' potential benefits in promoting healthier lifestyles and their risks of manipulation. The study, published on the preprint server arXiv as paper 2602.12924v1, explores the dual nature of GSAs as increasingly influential tools that can motivate users toward personal goals while simultaneously carrying potential risks of manipulation and deception. These risks stem from the limited control over probabilistic outputs that characterize current generative AI systems, particularly in healthcare contexts where patient motivation directly impacts treatment outcomes and recovery processes. According to the abstract, GSAs manifest communicative patterns based on the knowledge available to them, suggesting that their persuasive effectiveness is directly tied to the quality and scope of their training data and knowledge bases, making this research crucial for developing ethical and effective AI-assisted healthcare interventions.