Prompts para Interpretação Executiva pelo ChatGPT

This repository contains the experimental materials used in the study “Avaliação do uso de IA na geração de relatórios técnicos e executivos de desempenho”. The objective of this experiment is to evaluate how different prompt engineering strategies influence the generation of executive interpretations of software performance metrics using Large Language Models (LLMs). The study investigates whether an LLM can transform technical descriptions of performance metrics into clear, useful and accessible executive interpretations intended for managers and stakeholders without specialized knowledge in performance engineering. The experiment was conducted using GPT-5.5 through the ChatGPT platform under standard public access conditions available during the execution period. Three prompting strategies were evaluated: • Zero-Shot Prompting — task execution without examples or additional guidance. • Persona Prompting — execution guided by an explicitly defined expert role focused on software engineering, performance testing and executive communication. • Template Prompting — execution constrained by a predefined report structure. For each selected performance metric, the model receives: - metric name;- technical meaning of the metric. The generated outputs are intended to be evaluated through a survey with participants experienced in Software Engineering, Performance Testing, Software Quality, Performance Engineering or related areas. Evaluation criteria include: - clarity;- ease of understanding for non-technical managers;- interpretation accuracy;- usefulness for decision making;- adequacy to executive context. This repository currently provides: - prompt definitions used in the experiment;- survey template;- documentation required for reproducibility. Generated responses are not included in this version and may be released in future updates. Files included: - README.md- prompts.md- prompts.pdf