Work

Trust in AI decision support systems relates to innovation and technostress through empowerment and dependency pathways

Year: 2026

Type: article

Abstract: When employees trust AI-driven decision support systems (AI-DSS), does that trust help or hurt them? This study argues it does both at the same time. Drawing on Cognitive Appraisal Theory, Social Exchange Theory, and Technology Threat Avoidance Theory, we propose that trust in AI-DSS operates as a double-edged sword: it can free employees to think more creatively and generate new ideas (the empowerment pathway), yet it can also make them overly reliant on AI recommendations, creating stress and anxiety when the system changes or fails (the dependency pathway). Which effect dominates depends on two boundary conditions: employees with higher AI Literacy gain more from trust (innovation benefits), while stronger Organizational AI Support reduces the stress costs of dependency for everyone. We tested this model with a two-wave survey of 452 knowledge workers who use AI-DSS regularly. Results confirm both pathways: trust in AI-DSS significantly boosts user innovation behavior through perceived empowerment (indirect effect = 0.249), while simultaneously increasing technostress through perceived dependency (indirect effect = 0.175). AI Literacy amplifies the innovation benefit only above a critical threshold (scale score = 2.81), meaning that about one in four employees currently receives no innovation benefit from their trust. Organizational support, by contrast, reduces stress costs unconditionally, regardless of support level. A non-linear analysis further reveals an optimal trust zone: moderate-to-high trust maximizes net benefit, whereas excessive trust accelerates stress costs faster than it grows innovation returns. To estimate these effects robustly from Likert-scale survey data, we develop a Neural-Augmented Bayesian SEM (NB-SEM) framework that reduces estimation error by up to 51% compared with standard structural equation modeling. These findings suggest that the goal of AI-DSS governance should shift from simply building more trust to cultivating the right level of trust, supported by targeted AI literacy training and proactive organizational support structures.

Source: Scientific Reports

DOI: 10.1038/s41598-026-66908-0

Authors: Qijie Ruan, Xiaorui Han, Huinan Liu

Institutions: Chuzhou University, Hebei Finance University, Shanghai Zhongqiao Vocational and Technical University

Language: English

Topic: Technostress in Professional Settings

Subfield: Social Psychology

Field: Psychology

Domain: Social Sciences

SDG: Decent work and economic growth

Open Access status: gold

Estimated APC paid: $2,690