Misfortune or Blessing? The Effects of GAI Misinformation on Human-GAI Collaboration

Abstract

The advent of generative artificial intelligence (GAI) increasingly transforms artificial intelligence (AI) models from passive tools to proactive social actors. Organizations are increasingly deploying GAI as a teammate to improve work efficiency and decrease costs. However, this promising and sustainable development of GAI is uncertain because its output possesses much misinformation. Inadequate reliability and accuracy of GAI misinformation may distrust of human teammates and elicit unsatisfied collaboration outcomes. Decreasing over-reliance and over-trust may also offer the opportunity to dilute the harm of AI on human unique knowledge and activate innovative thoughts. Given these mixed theorizations, this paper leverages affordance actualization theory to draw a comprehensive understanding of misinformation in human-GAI team collaboration. Leveraging a mixed-methods design, we aim to begin with a qualitative study to identify the affordances of AI and interpret the actualization process. Based on the results, we intend to conduct a Wizard of Oz experiment to examine the proposed model.