Dark side of generative AI: a stressor–strain outcome and cognitive load theory explanation on decision fatigue and cognitive overload
Fahad Asmi, Alfred Wong, Intesar Almugren, Varun Chotia, Judit Bilinovics-SiposPurpose
To investigate when the use of generative AI (GenAI) in knowledge-intensive work remains sustainable vs when users intend to discontinue, this study models how multilevel stressors generate cognitive strains that drive discontinuance intentions. Drawing on the stress–strain–outcome (SSO) framework, supported by cognitive load theory (CLT), the study explains how task, technology and organizational demands translate into cognitive strain and discontinuance.
Design/methodology/approach
A cross-sectional survey of 460 knowledge workers was analyzed using covariance-based structural equation modeling (CB-SEM). The model incorporates task-level, technology-level and organizational-level stressors; identifies cognitive overload and decision fatigue as parallel strains; and specifies GenAI discontinuance intention as the outcome.
Findings
All stressors positively relate to cognitive overload and decision fatigue except content governance, which showed no significant effect on decision fatigue. Cognitive overload emerged as a significant factor while mapping decision fatigue, and both strains significantly contributed to GenAI Discontinuance Intention.
Practical implications
To achieve the sustainable use of GenAI, policymakers and facilitators should minimize task complexity, limit time-related pressures and implement measures that lower the risk of hallucination. They should also provide templates and establish standard operating procedures to minimize perceived overload and fatigue.
Originality/value
Extends the SSO framework for GenAI by integrating hallucination-related and governance stressors and by distinguishing dual cognitive strain pathways (cognitive overload, decision fatigue) leading to discontinuance.