Enhancing LLM Responses Through Instructional Design Theory: A Prompt-Based Approach

The fundamental function of an LLM is to generate textual responses to user queries that effectively support human tasks such as information acquisition, decision-making, and learning. For LLM responses to be useful, they must be accurate and logically sound. Consequently, extensive research has focused on improving these aspects. However, from a human-machine interaction perspective, usability is equally important. Beyond accuracy and logic, responses must be well-formatted and structured to facilitate human information processing. This critical aspect has received relatively little attention, representing a significant research gap. To address this, we proposed a response generation strategy grounded on instructional design theory to enhance the quality and structure of LLM responses. The strategy dynamically generates and updates the system prompt in real time to maintain alignment with instructional goals and users’ needs. Internal pilot testing suggests that this strategy improves the clarity and pedagogical alignment of responses in chatbot-based environments.