Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
How Can Teams Benefit From AI Team Members? Exploring the Effect of Generative AI on Decision-Making Processes and Decision Quality in Team–AI Collaboration
Citation. Zercher, D., Jussupow, E., Benke, I., & Heinzl, A. (2025). How Can Teams Benefit From AI Team Members? Exploring the Effect of Generative AI on Decision-Making Processes and Decision Quality in Team–AI Collaboration. Journal of Organizational Behavior. https://doi.org/10.1002/job.2898
Summary. This preregistered exploratory mixed-method laboratory experiment tested how an AI team member’s knowledge configuration changes team decision processes and quality in a hidden-profile task. Centralized AI knowledge reduced decision-making asymmetries and improved accuracy, whereas asymmetric AI knowledge did not outperform human-only teams. The authors argue that team–AI benefits depend on the alignment of AI knowledge with the team’s information distribution and on trust and information-processing processes.
Key claims
- Across 81 teams (215 participants), teams with centralized AI knowledge selected the correct candidate 67.85% of the time, versus 28.60% for teams with asymmetric AI knowledge and 28.00% for human-only teams; the centralized-AI condition differed significantly from both other conditions (χ²(2) = 11.77, p = 0.003, φ = 0.38).
- Centralized AI knowledge was associated with reduced reliance on prediscussion preferences and more beneficial AI information processing, while asymmetric AI knowledge produced only moderate reductions in decision-making asymmetries.
- Trust in the AI team member was lower when the AI had asymmetric rather than centralized knowledge (M = 3.21 vs. 3.47; U = 1696.50, p = 0.007), and the authors identify mistrust, nonbeneficial AI information processing, and a critical AI collaboration focus as AI-specific asymmetries that can prevent performance gains.
Evidence & limitations
- The study combines quantitative surveys with videotaped team discussions and ChatGPT-based interaction logs in a preregistered hidden-profile laboratory experiment.
- The findings identify mechanisms under controlled information configurations, but the task was a one-hour fictional airline-personnel-selection exercise with predominantly student participants, limiting direct generalization to sustained workplace teams and deployed AI systems.