"Prompt Engineering Versions Used for Bales IPA Classification"

"Abstract\u2014Automated analysis of group interaction in digital collaborative learning environments has become a strategic line of inquiry within the study of learning and educational technologies. The growing volume of discourse generated on digital platforms provides new opportunities to examine collective processes with greater scope and depth. In this context, advancing toward models that integrate scalability theoretical grounding, and interpretability constitutes a central priority for scientific development. Classical theoretical frameworks such as Bales\u2019 (1950) Interaction Process Analysis (IPA), widely established for the study of communicative dynamics in small groups, offer a promising pathway for adaptation to larger-scale digital environments.This study presents a hybrid methodological procedure that integrates Generative Artificial Intelligence (GAI) with the IPA model for the automated analysis of group interaction in digital educational contexts. The system combines an iterative process of prompt design and refinement with human supervision (human-in-the-loop), preserving the functional validity, categorical coherence, and analytical traceability of the original model while enabling the consistent processing of large discourse corpora. The approach is validated through the analysis of 7,321 contributions produced by 481 secondary school students (aged 15-16) during a structured online collaborative activity. The results reveal functional communication patterns consistent with the evolution of group Interaction during task resolution, reflecting transitions from individual reasoning to processes of deliberation, socio-emotional regulation, and the collective construction of consensus in shared decision-making.This work extends the applicability of the IPA model to large-scale digital contexts through the use of GAI and provides a transparent and replicable methodological infrastructure for the functional analysis of collaborative interaction. "