This study examines the persuasive effects of rewarding and punitive behaviors in social robots. In this study using the Pepper robot, we defined rewarding behavior as actions expressing positive empathy, and punitive behavior as actions expressing negative empathy. These definitions are specific to the behavior and capabilities of Pepper, but the concepts can be generalized to other social robots with similar functionalities. Our novel persuasion model, guided by the Elaboration Likelihood Model and balance theory, allows the robot to autonomously select behaviors. Experiments involving participants with strong negative attitudes towards robots demonstrated that a combination of rewarding and punitive strategies is more effective in eliciting desired behaviors than using only a positive attitude strategy alone. Results emphasize the importance of emotional expression, particularly tone of voice, in effective human-robot interaction.