2026-10-12T19:09:22Z https://www.scidb.cn/oai

10.57760/sciencedb.psych.000wy 2026-09-18T16:30:38Z

2026-09-18 The Impact of AI Engagement on Dyadic Team Decision-Making: Evidence from fNIRS-Based Inter-Brain Synchrony doi:10.57760/sciencedb.psych.000wy en Artificial intelligence (AI) is rapidly transitioning from a passive tool to an active collaborator, raising concerns about its impact on human learning and team coordination. However, the neural mechanisms underlying these effects, particularly in team contexts, remain largely unexplored. This study investigated how AI, as a team member, affects dyadic performance, learning, and inter-brain synchrony during a collaborative probabilistic reversal learning task, using functional near-infrared spectroscopy (fNIRS) hyperscanning. A total of 75 same-sex dyads of university students (N = 150) were randomly assigned to one of three groups: a dyad group (two human members collaborating), an AI-advice group (an additional AI advisor providing suggestions at 65% accuracy), or a historical-advice group (an additional suggestion from a previous human participant, identical in content to the AI advice). Behaviorally, we measured decision accuracy, trials to stable switching, learning slopes across task phases, and advice adoption. Neurally, we assessed regional brain activation, intra-brain functional connectivity (FC), and inter-brain synchronization (IBS) within the dorsolateral prefrontal cortex (DLPFC), frontopolar cortex (FPC), and temporo-parietal junction (TPJ). Behavioral results showed that the AI-advice group achieved the highest final decision accuracy and required the fewest trials to reach stable correct choices, particularly in early task stages. However, learning slope analysis revealed that only the dyad group exhibited a significant learning improvement trend across phases, indicating genuine strategy optimization over time. In advice adoption, participants consistently preferred human advice over AI advice. Notably, AI advice exerted a stronger influence on final decisions than historical advice, suggesting greater perceived competence of AI relative to an outside-human advisor. Furthermore, in the AI-advice group, the advising partner tended to choose options opposite to the previous AI suggestion, revealing a compensatory strategy to diversify information input. Neurally, both the AI-advice and historical-advice groups showed increased DLPFC activation compared to the dyad group, reflecting greater cognitive demands from integrating multiple information sources. Intra-brain functional connectivity analysis revealed that the AI-advice and dyad groups both exhibited stronger connectivity than the historical-advice group. Critically, the AI-advice group showed the lowest IBS in PFC and DLPFC regions, significantly lower than both the historical-advice and dyad groups, which did not differ from each other. This pattern indicates that the reduction in IBS is specifically driven by the “AI label” itself, i.e., AI as a non-social intelligent agent, rather than by the mere presence of an additional information source.These findings reveal a dissociation between performance and learning in AI-involved team collaboration: AI advice enhances immediate task performance, particularly in early stages, but does not facilitate, and may even impair sustained learning of task rules. Moreover, AI participation alters both individual cognitive processing and team-level neural coordination, as reflected in reduced inter-brain synchronization. By employing a three-group design that separates the AI label effect from general information-source effects, this study provides neural evidence that AI’s impact on team collaboration is uniquely driven by its perceived status as a non-social agent.  team collaboration; human-AI collaboration; task performance; cognitive processes cai hong hui PUBLIC https://creativecommons.org/licenses/by-nc-nd/4.0/ dataset Science Data Bank