How Can Teams Benefit From AI Team Members? – Research Article Published in the Journal of Organizational Behavior
In an exploratory experiment with 215 individuals in 81 teams, it was investigated how collaboration with generative AI team members can reduce decision-making asymmetries, and whether such collaboration introduces new, AI-specific asymmetries depending on the AI team members’ knowledge. Teams were asked to make decisions under distributed knowledge conditions, where human teams typically fail to make accurate decisions due to decision-making asymmetries, across three configurations of team-AI collaboration:
- Human-only teams with asymmetric knowledge
- Teams collaborating with AI team members with centralized knowledge
- Teams collaborating with AI team members with asymmetric knowledge
It was found that teams collaborating with AI team members with centralized knowledge made better decisions than human-only teams. This improvement was made possible by reduced decision asymmetries, trust in the AI, effective processing of the information provided by the AI, and constructive collaboration with the AI. In teams collaborating with AI team members with asymmetric knowledge, decision asymmetries were also reduced compared to human-only teams, but not as effectively as in teams collaborating with AI team members with centralized knowledge. Moreover, it was observed that certain AI-specific asymmetries emerged, including distrust in the AI, inefficient AI information processing, and a critical collaboration focus with the AI team member. Therefore, teams collaborating with AI team members with asymmetric knowledge were not able to outperform human-only teams.
The results illustrate the critical role of AI team members' knowledge and the importance of effective collaboration between human team members as well as between humans and AI team members for successful team-AI collaboration.
You can access the article under the following link: “https://onlinelibrary.wiley.com/doi/10.1002/job.2898” or download the article here.