Interactional Repair in Learner-AI Task-Based Conversations: Patterns of Initiation, Resolution and Participation (Replication Package)

This record provides the replication package for the study “Interactional Repair in Learner-AI Task-Based Conversations: Patterns of Initiation, Resolution and Participation.” The package includes: (1) analysis code and configuration files, (2) structured outputs used in reported findings, and (3) human validation templates and documentation. AI-assisted methods were used in two stages. OpenAI GPT-5 via the Custom GPT interface was used to develop the chatbot for data collection. OpenAI GPT-4o was used for candidate coding of repair sequences. All LLM outputs were reviewed and validated by human researchers. To protect participant privacy, public materials are de-identified and curated to exclude identifiable personal information. Funding: This work was supported by the Data Horizons Research Support Grants 2025, Data Horizons Research Centre, Macquarie University, Australia.