Divergence or Convergence? A Deep Insight into the Crowd Collaboration and its Productivity in Open Source Software based on Entropy

Paper Title

Divergence or Convergence? A Deep Insight into the Crowd Collaboration and its Productivity in Open Source Software based on Entropy

Publication Info

• Topic area: Open Source Software (OSS) collaboration and productivity analysis.
• Keywords: Open Source Software, fork practices, convergence entropy, productivity metrics, integration effectiveness, Jensen-Shannon divergence, distributed collaboration, project sustainability, issue resolution, commit activity.

Background and Problem

• Problem / challenge: Forking in OSS fosters innovation but often leads to inefficiencies, fragmentation, and challenges in integrating contributions back into the main repository. Existing metrics for integration effectiveness are static, coarse-grained, or focus on isolated inefficiencies, leaving the longitudinal relationship between integration effectiveness and project productivity underexplored.
• Significance: Effective integration of fork contributions is critical for OSS project sustainability, reducing bugs, improving productivity, and maintaining community cohesion.
• Motivation and related work: Prior studies have examined fork inefficiencies (e.g., lost contributions, duplicate pull requests) and proposed tools for improving integration pipelines. However, they lack a unified, time-sensitive measure of integration effectiveness and fail to explore its dynamic impact on productivity outcomes.

Solution

• Proposed approach: Introduction of Convergence Entropy, a novel metric based on Jensen-Shannon divergence, to quantify the integration effectiveness of fork practices in OSS repositories.
• Novelty:
  1. Development of convergence entropy to measure alignment between distributions of original and merged commits across forks, scaled by merge ratio.
  2. Application of convergence entropy to analyze its correlation with three productivity dimensions: new bugs, new commits, and issue resolution time.
  3. Identification of contextual factors (e.g., project type, age, number of forks) that modulate the relationship between convergence entropy and productivity.
• Procedure and key techniques:
  • Use Jensen-Shannon divergence to measure the similarity between distributions of original and merged commits across forks.
  • Scale this similarity by the ratio of merged to original commits to calculate convergence entropy.
  • Conduct panel regression analysis on eight prominent GitHub projects (four company-driven, four community-driven) to evaluate correlations with productivity metrics (new bugs, new commits, issue resolution time) while controlling for factors like project age, number of forks, issues, and contributors.

Results

• Concrete findings:
  • Convergence entropy is negatively correlated with new bugs, indicating improved software quality.
  • Correlation with new commits is positive in non-company projects but negative in company projects.
  • Correlation with issue resolution time is positive in non-company projects (prolonging resolution) and negative in company projects (accelerating resolution).
• Advantage over baselines: Convergence entropy provides a unified, time-sensitive measure of integration effectiveness, addressing limitations of prior metrics that were static, coarse-grained, or narrowly focused on specific inefficiencies.
• Experiments / evaluation:
  • Data collected from eight GitHub projects, encompassing 12,720 forks and 39,300,748 commits.
  • Regression models analyzed the impact of convergence entropy on productivity metrics, with interaction terms to explore moderating factors.
  • Results validated across company-driven and community-driven projects to account for organizational differences.
• Limitations and future work:
  • Dataset limited to eight large-scale projects, potentially limiting generalizability to smaller or inactive OSS projects.
  • Causal relationships remain speculative; future work will include longitudinal studies and experimental interventions.
  • Issue complexity, while controlled for statistically, may still influence findings.

Summary

This study introduces Convergence Entropy, a novel metric to quantify the integration effectiveness of fork practices in OSS projects. By analyzing its correlation with productivity metrics (new bugs, new commits, issue resolution time) across eight GitHub projects, the study reveals that convergence entropy enhances software quality, influences commit activity differently in company vs. non-company projects, and affects issue resolution efficiency based on project type. The findings highlight the socio-organizational dynamics of OSS collaboration and provide actionable recommendations for improving project governance and sustainability. Future work aims to expand the dataset and explore causal pathways to refine these insights.