Escaping the AI cage: construction workers’ perceived risk with AI gives rise to unsafe behaviors

Abstract

Reducing unsafe behavior among construction workers remains a central concern in construction management. While artificial intelligence (AI) technologies are increasingly adopted to replace or monitor workers, the negative perceptions workers hold toward the promotion of such technologies have largely been overlooked. This study introduces the concept of perceived risk with AI, defined as workers’ subjective perception of the potential negative consequences associated with AI adoption. Drawing on self-determination theory, which emphasizes individuals’ core psychological needs, we propose an inside-out theoretical model. This model explains how construction workers’ perceived risk with AI triggers unsafe behavior through two internal psychological pathways: emotional exhaustion and organizational identification. We further identify AI failure as a critical boundary condition that amplifies these effects. Data were collected from 343 frontline workers at 3 smart construction sites where AI actively interacts with human labor. Results indicate that perceived risk with AI increases unsafe behavior, and this relationship is mediated by both heightened emotional exhaustion and weakened organizational identification. Notably, these effects are significantly amplified when AI malfunctions occur. Our findings not only extend the application scope of self-determination theory but also contribute to the knowledge base on workers’ unsafe behavior in the context of human-AI coexistence. Practically, the study offers actionable insights for managers to mitigate the unintended side effects of AI implementation on worker safety.