Exploring Collaboration in Human-Artificial Intelligence Teams: A Design Science Approach to Team-AI Collaboration Systems

Research has long underscored the critical role of effective team collaboration in surpassing the limits of individual members’ capabilities. With organizations now increasingly integrating artificial intelligence (AI) as quasi-team members to enhance learning, problem-solving, and decision-making in teams, there is a pressing need to understand how to foster effective collaboration between teams and AI systems (i.e., team-AI collaboration). By adopting a design science approach and conducting nine semi-structured interviews with knowledge workers, we identify design requirements and principles for effective team-AI collaboration systems from an end-user perspective. We then develop a team-AI collaboration system within Discord (a voice, video, and text chat application) and evaluate its design through five laboratory experiments with human-AI teams. Our results show that introducing configurable roles and personalities for AI team members prompts humans to reconsider their own biases. However, human preconceptions still play a dominant role in shaping team performance.