Ensuring human agency: an interaction design pathway to AI augmentation

The augmentation of human work through Artificial Intelligence (AI) promises to be a panacea to the role of technology in organizations. While frameworks on augmentation theorize how to best divide work between humans and AI, the empirical literature on human-AI interaction offers unexpected and inconclusive findings. Interaction challenges—including overreliance and selected engagement with the algorithmic output—call into question how theorized augmentation benefits can be realized. Rooted in cognitive learning theory, this study’s conceptual model argues that human-AI interaction can lead to multiple beneficial outcomes when algorithmic output is designed in a reciprocal manner. By providing humans with reflection-provoking feedback, reciprocal algorithmic output does not prescribe any actions, and thereby necessitates humans to expend cognitive effort. Reciprocal algorithmic output enables three crucial augmentation outcomes: task performance, human agency, and human learning.