Increasingly intelligent AI artifacts in human-AI systems perform tasks more autonomously as entities that guide human actions, even changing the direction of task delegation between humans and AI. It has been shown that human-AI systems achieve better results when the AI artifact takes the leading role and delegates tasks to a human rather than the other way around.
Our study presents phenomena, conflicts, and challenges that arise in this process, explored through the theoretical lens of principal-agent theory (PAT). The findings are derived from a systematic literature review and an exploratory interview study and are placed in the context of existing constructs of PAT.
Furthermore, we identify new causes of tensions that arise specifically in AI-to-human delegation and calls for special mechanisms beyond the classical solutions of PAT. Our work thus contributes to the understanding of autonomous AI and its implications for human-AI delegation.
I am happy that our paper "Task delegation from AI to humans: A principal-agent perspective" has been accepted for presentation at the 44th International Conference on Information Systems (ICIS 2023). The conference will take place from December 10 to 13 in Hyderabad, India.