Agentic Relationship Dynamics in Human-AI Collaboration: A study of interactions with GPT-based agentic IS artifacts

Abstract

Generative Artificial Intelligence (AI) having become increasingly embedded into work in both academia and industry has put a magnifying glass on Human-AI collaboration. With this paper, we seek to answer calls for research on the interactions between human and AI agents and their outcomes. We adopt the IS Delegation Framework (Baird & Maruping, 2021) to look at dynamics in relationships between human agents and Generative Pre-trained Transformer-based agentic IS artifacts and how these dynamics manifest. By conducting and analyzing data from semi-structured interviews, we were able to identify five salient agentic relationship dynamics affecting common understanding, willingness to delegate, cognitive load in human agents, confidence, and human agents' abilities to break GPT-based agentic IS artifacts' "thought loops". With this, we aim to provide nuanced insight into GPT-based agentic IS artifacts and agentic relationship dynamics involving cognitive tasks.