Leveraging Prompt Engineering to Facilitate AI-driven Chart Generation

Within the field of artificial intelligence, persistent barriers to accurately understanding and representing digital data remain. This paper expounds on the importance of prompt engineering through a common numerical data interpretation problem in the current popular artificial intelligence models, and uses prompt engineering to solve the common shortcomings of the visual charts generated by the current artificial intelligence models. This study proposes a new protocol that aims to leverage prompt engineering to improve the accuracy of AI-driven chart generation. This paper used a dataset compiled from publicly available information and conducted my research using popular AI models such as ChatGPT 4o, Gemini Advanced, Claude Sonnet and Perplexity. The first round of tasks assigned to each model was to generate visualizations from the same data set without any specific prompts, and the four models produced completely different and meaningless visualizations. The second round of experiments followed, employing meticulously crafted prompts aimed at enhancing numerical comprehension and the precision of data representation, and the results indicated that all AI models exhibited enhanced comprehension of the data, as well as improved clarity and accuracy in the visualized graphs they generated. Furthermore, additional experiments were conducted to further demonstrate the efficacy of the protocol. The protocol not only enhances the models' data interpretation capabilities but also aligns their output with specific analytical goals, thus broadening their applicability in professional settings such as financial analysis, data analysis, and market research. By fostering a deeper integration of prompt engineering within AI systems, this research contributes to the development of more reliable and interpretable AI tools, ultimately supporting more informed decision-making in data-driven fields.