The emergence of Large Language Models (LLMs) has enhanced Social Robots (SRs), enabling more dynamic human interactions. Despite their growing use in healthcare, education, and customer service, their potential for persuasive communication remains underexplored. This research examines how tailored knowledge prompts shape the persuasive abilities of ChatGPT-powered SRs, focusing on the roles of assertiveness and expressiveness in Human-Robot Interaction (HRI).

This thesis applied the Knowledge-Behavior Model (KBM) and used a quantiative research approach. An online interaction study was conducted where participants solved decision tasks with resource allocations in scenarios like charitable donations and stress management. A robotic advisor tried to change participants resource allocation through persuasive messages. Persuasion effectiveness was measured based on participant's subsequent changes in resource allocation. In addition, participants rated their attitude towards the robot's persuasive behavior, and their atittude towards the robot.

The study found that context knowledge marginally significantly enhanced the attitude towards the robot and the attitude towards the robot's persuasive behavior. However, self knowledge did not yield significant results. In addition, the attitude towards the robot's persuasive behavior was significantly predicted by the attitude towards the robot. In turn, the attitude towards the robot's persuasive behavior significantly predicted persuasion effectiveness. These findings partially support the KBM model and highlight a potential effect of context knowledge on persuasion effectivenss, that is sequentially mediated by the attitude toward the robot and atittude towards the robot's persuasive behavior. These results provides valuable insights for deploying ChatGPT-powered SRs in practical applications, such as healthcare or customer service, where persuasive interactions are critical. By integrating advanced communication strategies, this thesis advances the understanding of ethical and effective AI-driven persuasion, offering a foundation for further studies on verbal and non-verbal communication in SRs.