The aging population and caregiver shortages pose major challenges for elderly care. Generative social robots (GSRs), powered by AI and large language models, offer potential solutions by encouraging healthy behaviors, supporting therapy, and providing emotional connection. However, their effectiveness in real-world caregiving remains underexplored, particularly regarding how robot knowledge influences persuasive interactions.

This study uses a two-phase methodology to explore how robot knowledge—self-knowledge, user personalization, and context-awareness—affects the persuasiveness of generative social robots (GSRs) in elderly care. In phase one, semi-structured interviews with caregivers identified key variables influencing effective interactions. In phase two, scenario-based simulations using ChatGPT tested different configurations of robot knowledge to evaluate their persuasiveness based on different knowledge types.

The study reveals that user-knowledge and context-knowledge may be the most prominent factors influencing the persuasiveness of generative social robots (GSRs) in elderly care. User-knowledge allows for personalized interactions, leveraging biographical and emotional awareness to create meaningful connections. Context-knowledge enhances situational relevance by aligning robot actions with broader user goals and caregiving environments. Self-knowledge, while foundational for fostering trust and maintaining professionalism, contributes less to overall persuasiveness compared to the other dimensions. Advanced configurations that integrate all three knowledge areas demonstrated a plateau effect, indicating diminishing returns when all dimensions are combined. The findings highlight that user- and context-awareness alone are sufficient for effective and engaging interactions, while self-knowledge ensures a baseline of empathy and respect. The study also emphasizes the importance of balancing personalization and autonomy with assertiveness to maximize the robot's motivational impact.