AI innovation suppression in the context of employee–AI–colleague power dynamics

This study draws on power–dependence theory and integrates three scenario–based experiments and a three–wave survey analysis involving 475 questionnaires. The findingsreveal: (1) As we shift our understanding of human–AI relationships from a tool–based view to oneof dependence on human–like agents, intelligence in AI systems shows a significant negativecorrelation with employees’ innovative behavior; (2) as a human–like agent, AI suppressesinnovation by reducing their sense of power; (3) the mediating role of employees’ sense of powerbetween intelligence in AI systems and innovative behavior is differentially moderated by the typeof helping behaviors between employees and colleagues: helping–giving weakens this mediation,whereas helping–receiving strengthens it.

This research moves beyond the existing dyadicframework of human–AI interaction by constructing a triadic power architecture of “employee–AI–colleague.” This study not only identifies the empowerment paradox of intelligence in AI systemsbut also expands theoretical understanding of the mechanisms and boundary conditions underlyinginnovation suppression. The findings provide evidence–based insights for enterprises to designthresholds for intelligence in AI systems, establish a balanced framework between “technologicaldeployment and psychological empowerment,” and build compensatory organizational supportsystems, thus offering practical guidance for addressing the innovation–inhibiting dilemma causedby intelligent systems.