Abstract

The rapid expansion of artificial intelligence applications in conversational systems has introduced new challenges in anticipating user needs and generating responses that are contextually appropriate across multiple turns of dialogue. A novel approach was introduced to address these challenges, incorporating rehearsed prompt generation combined with perspective-taking strategies to enable AI models to anticipate follow-up questions before they are posed. Through systematic prompt rehearsals, GPT-Neo was trained to simulate various user viewpoints and generate more adaptive responses, thereby improving the model's capacity to manage complex, multi-turn interactions. The experimental results revealed substantial improvements in coherence, relevance, and diversity, particularly when rehearsed prompts were combined with perspective-taking techniques. Additionally, automated evaluation metrics confirmed the model's enhanced performance across multiple domains, suggesting significant applications in customer service, educational tutoring, and research assistance. These findings reaffirm the practical value of anticipatory question answering in AI systems, offering a scalable and efficient solution for improving user experience in dynamic environments.