Solaria-GPT 🧠🛠️

A Tailored ChatGPT for Usability Inspection

Solaria-GPT is an AI-driven tool designed to assist in usability inspection by detecting defects and classifying heuristic violations. Built on top of ChatGPT, the system supports two types of input formats:

• 📝 Text Input — A textual description of a usability defect.
• 🖼️ Media Input — A screenshot of the system interface or a video showcasing user interaction/navigation.

---

🔧 How to Create a Tailored ChatGPT

✅ Prerequisites

• Active ChatGPT Plus subscription.

🧭 Step-by-Step Guide

1. Access the Explore GPTs section.

2. Click on the "Create" button.

3. Use the "Create" tab to configure your GPT via conversational prompts.

4. Use the "Configure" tab to fill in model details and capabilities.

   • Name: "Solaria-GPT"
   • Description: "Solaria-GPT is an assistant that detects usability issues and classifies them according to Nielsen's heuristics. To use it, provide a textual description of the problem or a screenshot/video of the design to be evaluated."
   • Instructions: Copy and Paste the Instructional Prompt for Solaria-GPT in the field.
   • Conversation Startes: Copy and Paste the following icebreakers:

     Which heuristic is violated in this defect description?
     

     What are the usability issues in the attached video?
     

     Analyze the attached video and perform a usability inspection.
     

     Are there any visible usability issues in the provided image?
     

   • Knowledge: Attach the file Nielsen's Heuristics File with Examples in the field.

5. Share your tailored GPT with others by generating a public link.

We illustrated the process to tailor a ChatGPT below:

---

🚀 Getting Started

How to use Solaria-GPT?

1. Access Solaria-GPT: https://chatgpt.com/g/g-gO0eGq3rN-solaria-gpt
2. Data Input: Enter some input data
   • Text: description of some usability defect
   • screenshot: a screenshot of the software that you want to check for usability defects
   • video: a recording of a video interaction with the software that you want to check for usability defects
3. Processing: Wait for the Solaria-GPT to process your input
4. Response: Solaria-GPT will have the following response:
   • For text: the violated heuristic by the problem inserted
   • For media (screenshot and video): the defect description, the location where the problem happens, and the violated heuristic

---

🧠⚙️ Prompt Instructions

• 📜 Instructional Prompt for Solaria-GPT
• 📘 Nielsen's Heuristics File with Examples

---

Experiments and Comparison with Other Tools

📦 Datasets

• 📁 Defect Database
• 🖼️ Screenshot Collection
• 🎥 Video Samples

---

🤖⚙️ Prompt Strategies

• 📑 Prompt Examples Used with Solaria-GPT:
  • Text:

    Which heuristic is violated in the following defect.
    

  • Screenshot:

    Perform a usability inspection of the attached screen.
    

  • Video:

    Analyze the attached video and perform a usability inspection.
    

---

💻 Execution

• Use the dataset available
• Use the prompt strategies
• TOOL must be: Solaria-GPT, Qween, DeepSeek, Claude, and Gemini.

Textual Validation

1. Open the [TOOL]
2. Type the prompt strategy for textual input
3. Add the defect from the defects database

Screenshot Validation

1. Open the [TOOL]
2. Type the prompt strategy for screenshot input
3. Attach the screenshot from the screenshot collection

Video Validation

1. Open the [TOOL]
2. Type the prompt strategy for video input
3. Attach the video from the video samples

---

📊 Output & Results

• 📂 heuristic_classification_results.xlsx — Spreadsheet with classification outcomes
• 📂 screenshot_defect_detection.xlsx — Results from screenshot analysis
• 📂 video_defect_detection.xlsx — Results from video analysis

---

📄 License

This project is licensed under the MIT License.

---

📝Paper Reference

Chaves, L., Lima, M., & Conte, T. (2025). Solaria-GPT: A Tailored ChatGPT Tool for Usability Inspection. In Proceedings of the XXXIX Brazilian Symposium on Software Engineering (SBES 25). Recife, PE, Brazil, 2025. Pre-print.

👥 Authors

This project was developed by:

• Lennon Correa Chaves – lennon@icomp.ufam.edu.br
• Márcia Sampaio Lima – msllima@uea.edu.br
• Tayana Uchoa Conte – tayana@icomp.ufam.edu.br

🏛️ Institutions:

• Institute of Computing - Federal University of Amazonas - UFAM. 🔗 https://icomp.ufam.edu.br/
• State University of Amazonas - UEA. 🔗https://www.uea.edu.br/
• USES Research Group. 🔗https://uses.icomp.ufam.edu.br/

📚 Citation

If you use Solaria-GPT artifact in your research or project, please cite it as follows:

Chaves, L., Lima, M., & Conte, T. (2025). Supplementary Material - Solaria-GPT: A Tailored ChatGPT Tool for Usability Inspection. In Proceedings of the XXXIX Brazilian Symposium on Software Engineering (SBES 25). Recife, PE, Brazil, 2025. DOI: 10.5281/zenodo.15282435.

@misc{solariaGPT,
  author       = {Lennon Chaves, Marcia Lima, Tayana Conte},
  title        = {{Supplementary Material - Solaria-GPT: A Tailored ChatGPT Tool for Usability Inspection}},
  year         = {2025},
  howpublished = {\url{https://github.com/lennonchaves/Solaria-GPT}},
  doi = {10.5281/zenodo.15282435}
}

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request for suggestions and improvements.