Presentation
Working With or Around You: The Impact of Imperfect Cooperation with an AI Agent on Team Performance and Teamwork
SessionPoster Session 1
DescriptionAdvances in AI presage effective human-machine teaming (HMT), but teamwork is complex and not assured, even if AI is trained to prioritize such behavior. This experiment explored how manipulations of an AI agent’s ability to cooperate with its human teammate and insight into the agent’s decision making influenced team performance, human teammate workload, and beliefs about the agent. Participants played a dyadic, online version of the board game Pandemic with a custom AI agent. The agent’s ability to cooperate with its human teammate and the availability of a display that gave insight into the agent’s decision making were manipulated between participants. Analysis of our data indicated that all teams won a similar number of games across experimentally manipulated conditions, suggesting that participants were able to work effectively with the AI agent. However, participants assigned to work with the less cooperative agent initially took more actions to win each game, rated their workload as higher, and reported they understood the agent less, relative to participants working with the more cooperative agent. The transparency display manipulation did not ameliorate these effects. AI designers are cautioned to carefully consider how cooperation with human teammates will be enabled to support team performance in HMT.
Contributors
Event Type
Poster
TimeTuesday, October 14th5:30pm - 6:30pm CDT
LocationRiverside East
Similar Presentations


