Agile Retrospectives: What went well? What didn't go well? What should we do?
Maria Spichkova, Hina Lee, Kevin Iwan, Madeleine Zwart, Yuwon Yoon, Xiaohan Qin·April 16, 2025
Summary
RetroAI++ is a web-based tool designed for Agile/Scrum retrospectives, focusing on enhancing team learning and anonymity. It uses Large Language Models to anonymize comments, improve team safety, and efficiency. The tool simplifies retro-meetings and provides data-based inputs. Experiments compare human vs. machine categorization accuracy using OpenAI's ChatGPT-4 turbo. Future work aims to expand experiments on a larger dataset and refine the tool.
Introduction
Background
Overview of Agile/Scrum methodologies
Importance of retrospectives in Agile/Scrum
Challenges in traditional retrospectives
Objective
Aim of RetroAI++
Key features and benefits
Method
Data Collection
Gathering user feedback and comments
Selection of data for experiments
Data Preprocessing
Cleaning and anonymizing data
Preparation for machine learning models
Model Training
Utilization of Large Language Models (LLMs)
Comparison of human vs. machine categorization
Use of OpenAI's ChatGPT-4 turbo
Evaluation
Metrics for assessing accuracy
Analysis of categorization results
Results
Categorization Accuracy
Comparison of human and machine categorization
Insights into model performance
Team Safety and Anonymity
Impact on team dynamics
Feedback on anonymity enhancement
Future Work
Expanding Experiments
Larger dataset for more robust results
Diverse user groups for broader applicability
Tool Refinement
Iterative improvements based on feedback
Integration of new LLM advancements
Conclusion
Summary of Findings
Key outcomes of the study
Implications for Agile/Scrum Teams
Practical applications of RetroAI++
Recommendations for future use
Outlook
Potential future developments in LLMs and Agile/Scrum tools
Basic info
papers
software engineering
artificial intelligence
Advanced features
Insights
How does RetroAI++ utilize Large Language Models to enhance anonymity in Agile/Scrum retrospectives?
In what ways does RetroAI++ integrate with existing Agile/Scrum tools to streamline retrospective meetings?
What are the key features of RetroAI++ that contribute to improving team safety and efficiency?
What are the proposed future enhancements for RetroAI++ to improve its performance on larger datasets?