From Prompts to Performance: Leveraging LLMs for Enhanced Educational AI Interactions

The rise in usage of Artificial Intelligence (AI) tools such as ChatGPT emphasizes the significance of exploring how prompt inputs influence response quality. This research analyzed 308 prompt-response pairs sourced from students taking Project-based classes like AI, machine learning, and Game design. Prompt quality was assessed based on specificity, clarity, and tone, while response quality was measured based on accuracy, relevance and complexity using a 5-point Likert scale. Our study explored correlation between various factors such as prompt specificity, response accuracy and relevance using Pearson analysis. Furthermore, we evaluated the performance of ChatGPT across each prompt type to determine which prompt class generated the best and worst results. Through the adoption of a prompt engineering approach, the paper identifies practical recommendations that students can employ to improve their interactions with ChatGPT. The research emphasizes that advanced methods such as prompt chaining and knowledge-augmented prompts are more effective than simple prompts in generating precise and comprehensive outputs successfully. This study shows ChatGPT's performance across various domains. The outcome of this study aims to help educators and learners to understand how to use generative AI effectively for programming problems and to support the development of efficient AI -enhanced learning environments.