The effects of generative AI usage on employee knowledge behavior: a perspective based on the Automation, Augmentation Paradox

Generative AI is reshaping knowledge work, yet its influence on employee knowledge behavior remains theoretically fragmented. Drawing on the Automation, Augmentation Paradox and Cognitive Appraisal Theory, this study constructs a dual-pathway moderated mediation model to examine how generative AI usage simultaneously affects knowledge sharing and knowledge hiding. We propose that AI usage enhances self-efficacy through the augmentation mechanism, thereby promoting knowledge sharing through the empowerment pathway, while simultaneously heightening job insecurity through the automation mechanism, thereby reinforcing knowledge hiding through the threat pathway. To capture the net behavioral tendency, we introduce Knowledge Behavior Relative Intensity, defined as the difference between knowledge sharing and knowledge hiding scores, as an integrative outcome variable. Furthermore, competitive psychological climate is examined as a moderator that amplifies both pathways. Using survey data from 428 knowledge workers in China, we tested the hypotheses with hierarchical regression and PROCESS bootstrap analyses. Results supported all hypotheses: AI usage positively predicted both knowledge sharing and knowledge hiding, with the former effect substantially stronger. The two indirect effects constituted competitive mediation, with opposing directions that statistically offset each other. Competitive psychological climate simultaneously strengthened both pathways. These findings advance the understanding of AI's paradoxical effects on knowledge behavior and offer practical implications for organizations managing AI adoption alongside knowledge governance.