Abstract—Team conflict within organizations is a vital predictor of project outcomes in today’s organizations, especially in digitally mediated settings where interaction mostly happens through text-based communication. Determining the conflict state within projects and predicting project outcomes based on these interactions is a difficult task. Currently, there exist no methods that use automated analyses, dynamic linguistic features, and multivariate prediction models to understand behavioral dynamics, temporal relations, and cross-variables associations between conflict states and project outcomes. Moreover, many available datasets are not annotated for conflict and performance states and cannot be used for supervised learning methods. In order to tackle these issues, this paper suggests a new behavioral intelligence system based on a Large Language Model (LLM) to predict team conflicts and project outcomes from meetings’ transcripts. This paper proposes an approach that incorporates an easily interpretable proxy labeling scheme for the generation of conflict and project outcome labels based on measurable signs of conversation, including sentiment polarity changes, interruptions, speaker dominance, and agenda management actions. Linguistic, interactional, and contextual cues are combined with transformer contextual embedding produced by a shared DeBERTa encoder. Multi-task learning is used to benefit from the common representation of the behavior across the two prediction tasks. Experiments show that the model reaches high levels of predictive power, with macro-average precision, recall, and F1-Scores of 91%, 89%, and 90%, respectively, on predicting conflict intensity and accuracy of 93% with a macroROC-AUC of 95% on predicting project outcomes. This approach is implemented using Python and PyTorch.
 
Keywords—team conflict detection, project outcome prediction, behavioral intelligence, large language models, multi-task learning
 
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).