Unpacking Human-AI Collaboration: The Interplay of Ability, Task Conflict, and Flow via NLP-Based Behavioral Analytics

As artificial intelligence (AI) becomes an integral member of collaborative teams, understanding the differences in performance and willingness to collaborate between human-AI and human-human teams is increasingly important. Drawing on social comparison and social motivation theories, this study integrates natural language processing (NLP) techniques to systematically examine how collaboration type (human-AI vs. human-human), task conflict, and flow experience influence team performance and subsequent collaboration intentions, with individual ability as a moderator. Based on an experiment with 181 university students and the collection of open-ended textual responses, we employed NLP methods such as BERT embeddings and principal component analysis to quantitatively link textual features with psychological variables. The results show that human-AI collaboration enhances both performance and willingness to collaborate again, and NLP analyses reveal significant associations between participants’ textual content and their experiences of task conflict and flow. This research provides a data-driven perspective for optimizing intelligent teamwork and understanding human-AI collaboration mechanisms.