The Impact of ChatGPT Utilization on the Co-Evolution Process in Multidisciplinary Collaboration: Changes in Conflict Management Behavior

Abstract

Background Modern society increasingly demands integrated approaches across disciplines to solve complex problems, thereby highlighting the growing importance of multidisciplinary collaboration. Recent advances in generative artificial intelligence (AI), with ChatGPT as a representative example, have gained attention as supportive tools in such collaborations. However, the specific roles of ChatGPT in multidisciplinary settings and the qualitative changes it induces in collaboration processes remain insufficiently explored. Grounded in co-evolution theory and conflict management theory, this study aims to empirically analyze how ChatGPT facilitates interactions between problem definition and solution generation, and how it influences affective process and cognitive conflicts within multidisciplinary collaboration. Methods To compare the effects of ChatGPT usage in multidisciplinary collaboration, six co-creation workshops were conducted. Groups were divided into an experimental group using ChatGPT and a control group without ChatGPT. Protocol analysis was used to measure the frequency of co-evolution components: problem (P), solution (S), affective conflict (A), process conflict (PR), and cognitive conflict (C). Co-evolution graphs were constructed based on co-evolution cycles (t, T), repetition frequency, and extinction (v), and independent sample t-tests were conducted to analyze group differences. Pre- and post-workshop surveys measured changes in conflict-handling styles (competing and compromising), and post-workshop semi-structured interviews qualitatively analyzed ChatGPT's roles as a mediator (m) and facilitator (f). Results The ChatGPT-supported group exhibited significantly higher interaction frequencies between problems and solutions, and greater numbers of co-evolution cycles than the control group. Independent t-tests showed statistically significant differences in P and S ratios (p < .05), and the average number of co-evolution cycles (T) in the experimental group was over three times higher than that of the control group. Affective (A) and process (PR) conflicts were significantly lower in the ChatGPT group, while cognitive conflict (C) was more active, promoting creative problem-solving. More than 75% of participants showed increases in competing and compromising behavior scores, suggesting that ChatGPT influenced conflict behavior during collaboration. Interview results revealed that ChatGPT helped to clarify vague ideas, to present diverse perspectives, and to deepen discussions. ChatGPT also acted as a psychological mechanism that provided logical grounds and trust, strengthening participants' arguments and mitigating conflicts. Conclusions This study empirically demonstrates that ChatGPT plays dual roles as a mediator and facilitator in multidisciplinary collaboration. ChatGPT enhances the co-evolution process by increasing iterative cycles and improving completeness, alleviating negative conflicts (A, PR), and fostering positive cognitive conflict (C) that drives creativity. By reinforcing the cyclical interaction between problem definition (Focus) and solution development (Fitness), ChatGPT contributes to the core characteristics of co-evolution in multidisciplinary work. Nonetheless, limitations are observed in its ability to integrate complex ideas and to guide discussion direction, indicating the continued need for complementary human collaboration. This study provides both theoretical and practical implications for the design of AI-supported collaboration by examining the potential and limitations of generative AI in real-world settings.