Uso do ChatGPT para interpretar métricas de desempenho e gerar relatórios executivos, avaliando diferentes estratégias de prompting

This repository contains the experimental materials used in the study "Uso do ChatGPT para interpretar métricas de desempenho e gerar relatórios executivos, avaliando diferentes estratégias de prompting". The objective of this experiment is to evaluate how different prompt construction strategies influence the ability of a Large Language Model (LLM) to transform technical software performance information into executive interpretations understandable by managers and stakeholders without expertise in performance engineering. The study evaluates the GPT-5.5 model through the ChatGPT platform using three prompting strategies: Zero-Shot Prompting; Persona Prompting; Template Prompting. Each execution received the following inputs: Context of the evaluated system; Performance metric name; An image containing performance test results. The experimental context was provided before all executions to standardize the generated interpretations. The generated outputs were anonymized and evaluated through a survey with non-specialist participants using a 5-point Likert scale considering: Information relevance; Comprehensibility; Explanatory capability; Usefulness for decision-making; Suitability for an executive audience. This repository includes the complete materials required for reproducibility and methodological inspection of the study: Experimental context (context/context.txt); Prompt definitions (prompts/*.txt); Survey instrument (survey/survey.pdf); Anonymized survey responses (survey/responses.csv). Model configuration: Model: GPT-5.5 Platform: ChatGPT Access type: Free account Configuration: Default platform settings Execution period: June 2026 License: CC BY 4.0https://creativecommons.org/licenses/by/4.0/